Text-guided image inpainting endeavors to generate new content within specified regions of images using textual prompts from users. The primary challenge is to accurately align the inpainted areas with the user-provided prompts while maintaining a high degree of visual fidelity. While existing inpainting methods have produced visually convincing results by leveraging the pre-trained text-to-image diffusion models, they still struggle to uphold both prompt alignment and visual rationality simultaneously. In this work, we introduce FreeInpaint, a plug-and-play tuning-free approach that directly optimizes the diffusion latents on the fly during inference to improve the faithfulness of the generated images. Technically, we introduce a prior-guided noise optimization method that steers model attention towards valid inpainting regions by optimizing the initial noise. Furthermore, we meticulously design a composite guidance objective tailored specifically for the inpainting task. This objective efficiently directs the denoising process, enhancing prompt alignment and visual rationality by optimizing intermediate latents at each step. Through extensive experiments involving various inpainting diffusion models and evaluation metrics, we demonstrate the effectiveness and robustness of our proposed FreeInpaint.

Recent advances in text-to-video generation have demonstrated the substantial superiority of diffusion models. Nevertheless, generating high-resolution videos based on text description still faces a great challenge due to the enormous computation overhead for video diffusion model training. In this paper, we present a tuning-free video diffusion approach with Spatial-Temporal LAtent Grouping (ST-LAG), for highresolution video generation. ST-LAG exploits the prior knowledge of a pre-trained low-resolution video diffusion model for regionwise video latent denoising, and then combines all the denoised regions of video latent as a whole one to achieve global-wise spatial-temporal coherence. Specifically, ST-LAG denoises the whole video latents via two deliberately designed modules, e.g., Spatial Latent Grouping (SLG) and Temporal Latent Grouping (TLG), at spatial and temporal level, respectively. SLG spatially slices the latent of each frame into different local patches, and then feeds them into the low-resolution video diffusion model for local-region latent denoising. A text re-weighting scheme is devised in SLG to strength the cross-attention between features of text tokens and spatial regions to facilitate spatial-level finegrained details generation. TLG capitalizes on the segmentlevel latent grouping to match the length of each denoised local segment with the frame number in the training stage. The well aligned temporal receptive field facilitates better preservation of motion patterns. In each denoising step, all groups of video latent at spatial and temporal levels are fused together for highresolution video generation. Extensive experiments conducted on the ECTV-Prompt dataset demonstrate the effectiveness of our approach quantitatively and qualitatively.

Recent advances in score distillation sampling (SDS) have revolutionized the field of text-to-3D generation, enabling the distillation of prior knowledge from diffusion models for 3D generation. Although exhibiting impressive texture quality, these methods often suffer from geometric inconsistencies (“Janus” issue), as the prior 2D diffusion model inherently lacks 3D awareness. Recent work fine-tunes the 2D diffusion model on 3D data to obtain a multi-view diffusion model as the SDS prior, which addresses the Janus issue but is at the cost of sacrificing texture quality, as available 3D training data always have unrealistic texture. Thus, a natural question arises — Is therean ideal prior diffusion model for 3D generationthat simultaneously has 3D awareness and high texture fidelity? In response, we present HiFi3D, a tuning-free method to establish a new hybrid diffusion model that can generate consistent multi-view images with photorealism appearances. We accomplish this by novelly marring a 3D multi-view diffusion model with a 2D image diffusion model through our unique designs. We find that such a high-fidelity multi-view diffusion model harbors an innate agency to serve as a strong prior for SDS optimization. Additionally, we introduce a depth-guided multi-view attention strategy to further improve the 3D consistency across views during optimization. Extensive experiments demonstrate that our HiFi3D outperforms previous state-of-the-art methods in faithfully generating 3D content with realistic textual details and consistent geometry.

Perpetual view generation aims to synthesize a long-term video corresponding to an arbitrary camera trajectory solely from a single input image. Recent methods commonly utilize a pre-trained text-to-image diffusion model to synthesize new content of previously unseen regions along camera movement. However, the underlying 2D diffusion model lacks 3D awareness and results in distorted artifacts. Moreover, they are limited to generating views of static 3D scenes, neglecting to capture object movements within the dynamic 4D world. To alleviate these issues, we present DreamJourney, a two-stage framework that leverages the world simulation capacity of video diffusion models to trigger a new perpetual scene view generation task with both camera movements and object dynamics. Specifically, in stage I, DreamJourney first lifts the input image to 3D point cloud and renders a sequence of partial images from a specific camera trajectory. A video diffusion model is then utilized as generative prior to complete the missing regions and enhance visual coherence across the sequence, producing a cross-view consistent video adheres to the 3D scene and camera trajectory. Meanwhile, we introduce two simple yet effective strategies (early stopping and view padding) to further stabilize the generation process and improve visual quality. Next, in stage II, DreamJourney leverages a multimodal large language model to produce a text prompt describing object movements in current view, and uses video diffusion model to animate current view with object movements. Stage I and II are repeated recurrently, enabling perpetual dynamic scene view generation. Extensive experiments demonstrate the superiority of our DreamJourney over state-of-the-art methods both quantitatively and qualitatively. Our project page:https://dream-journey.vercel.app/.

The first-in-first-out (FIFO) video diffusion, built on a pre-trained text-to-video model, has recently emerged as an effective approach for tuning-free long video generation. This technique maintains a queue of video frames with progressively increasing noise, continuously producing clean frames at the queue's head while Gaussian noise is enqueued at the tail. However, FIFO-Diffusion often struggles to keep long-range temporal consistency in the generated videos due to the lack of correspondence modeling across frames. In this paper, we propose Ouroboros-Diffusion, a novel video denoising framework designed to enhance structural and content (subject) consistency, enabling the generation of consistent videos of arbitrary length. Specifically, we introduce a new latent sampling technique at the queue tail to improve structural consistency, ensuring perceptually smooth transitions among frames. To enhance subject consistency, we devise a Subject-Aware Cross-Frame Attention (SACFA) mechanism, which aligns subjects across frames within short segments to achieve better visual coherence. Furthermore, we introduce self-recurrent guidance. This technique leverages information from all previous cleaner frames at the front of the queue to guide the denoising of noisier frames at the end, fostering rich and contextual global information interaction. Extensive experiments of long video generation on the VBench benchmark demonstrate the superiority of our Ouroboros-Diffusion, particularly in terms of subject consistency, motion smoothness, and temporal consistency.

Video virtual try-on aims to seamlessly dress a subject in a video with a specific garment. The primary challenge involves preserving the visual authenticity of the garment while dynamically adapting to the pose and physique of the subject. While existing methods have predominantly focused on image-based virtual try-on, extending these techniques directly to videos often results in temporal inconsistencies. Most current video virtual try-on approaches alleviate this challenge by incorporating temporal modules, yet still overlook the critical spatiotemporal pose interactions between human and garment. Effective pose interactions in videos should not only consider spatial alignment between human and garment poses in each frame but also account for the temporal dynamics of human poses throughout the entire video. With such motivation, we propose a new framework, namely Dynamic Pose Interaction Diffusion Models (DPIDM), to leverage diffusion models to delve into dynamic pose interactions for video virtual try-on. Technically, DPIDM introduces a skeleton-based pose adapter to integrate synchronized human and garment poses into the denoising network. A hierarchical attention module is then exquisitely designed to model intra-frame human-garment pose interactions and long-term human pose dynamics across frames through pose-aware spatial and temporal attention mechanisms. Moreover, DPIDM capitalizes on a temporal regularized attention loss between consecutive frames to enhance temporal consistency. Extensive experiments conducted on VITON-HD, VVT and ViViD datasets demonstrate the superiority of our DPIDM against the baseline methods. Notably, DPIDM achieves VFID score of 0.506 on VVT dataset, leading to 60.5% improvement over the state-of-the-art GPD-VVTO approach.

Animating images with interactive motion control has garnered popularity for image-to-video (I2V) generation. Modern approaches typically rely on large Gaussian kernels to extend motion trajectories as condition without explicitly defining movement region, leading to coarse motion control and failing to disentangle object and camera moving. To alleviate these, we present MotionPro, a precise motion controller that novelly leverages region-wise trajectory and motion mask to regulate fine-grained motion synthesis and identify target motion category (i.e., object or camera moving), respectively. Technically, MotionPro first estimates the flow maps on each training video via a tracking model, and then samples the region-wise trajectories to simulate inference scenario. Instead of extending flow through large Gaussian kernels, our region-wise trajectory approach enables more precise control by directly utilizing trajectories within local regions, thereby effectively characterizing fine-grained movements. A motion mask is simultaneously derived from the predicted flow maps to capture the holistic motion dynamics of the movement regions. To pursue natural motion control, MotionPro further strengthens video denoising by incorporating both region-wise trajectories and motion mask through feature modulation. More remarkably, we meticulously construct a benchmark, i.e., MC-Bench, with 1.1K user-annotated image-trajectory pairs, for the evaluation of both fine-grained and object-level I2V motion control. Extensive experiments conducted on WebVid-10M and MC-Bench demonstrate the effectiveness of MotionPro. Please refer to our project page for more results: https://zhw-zhang.github.io/MotionPro-page/.

Diffusion models have shown preliminary success in virtual try-on (VTON) task. The typical dual-branch architecture comprises two UNets for implicit garment deformation and synthesized image generation respectively, and has emerged as the recipe for VTON task. Nevertheless, the problem remains challenging to preserve the shape and every detail of the given garment due to the intrinsic stochasticity of diffusion model. To alleviate this issue, we novelly propose to explicitly capitalize on visual correspondence as the prior to tame diffusion process instead of simply feeding the whole garment into UNet as the appearance reference. Specifically, we interpret the fine-grained appearance and texture details as a set of structured semantic points, and match the semantic points rooted in garment to the ones over target person through local flow warping. Such 2D points are then augmented into 3D-aware cues with depth/normal map of target person. The correspondence mimics the way of putting clothing on human body and the 3D-aware cues act as semantic point matching to supervise diffusion model training. A point-focused diffusion loss is further devised to fully take the advantage of semantic point matching. Extensive experiments demonstrate strong garment detail preservation of our approach, evidenced by state-of-the-art VTON performances on both VITON-HD and DressCode datasets. Code is publicly available at: https://github.com/HiDream-ai/SPM-Diff.

Autoregressive models have emerged as a powerful generative paradigm for visual generation. The current de-facto standard of next token prediction commonly operates over a single-scale sequence of dense image tokens, and is incapable of utilizing global context especially for early tokens prediction. In this paper, we introduce a new autoregressive design to model a hierarchy from a few low-resolution image tokens to the typical dense image tokens, and delve into a thorough hierarchical dependency across multi-scale image tokens. Technically, we present a Hierarchical Masked Autoregressive models (Hi-MAR) that pivot on low-resolution image tokens to trigger hierarchical autoregressive modeling in a multi-phase manner. Hi-MAR learns to predict a few image tokens in low resolution, functioning as intermediary pivots to reflect global structure, in the first phase. Such pivots act as the additional guidance to strengthen the next autoregressive modeling phase by shaping global structural awareness of typical dense image tokens. A new Diffusion Transformer head is further devised to amplify the global context among all tokens for mask token prediction. Extensive evaluations on both class-conditional and text-to-image generation tasks demonstrate that Hi-MAR outperforms typical AR baselines, while requiring fewer computational costs.

Recent advancements in image generative foundation models have prioritized quality improvements but often at the cost of increased computational complexity and inference latency. To address this critical trade-off, we introduce HiDream-I1, a new open-source image generative foundation model with 17B parameters that achieves state-of-the-art image generation quality within seconds. HiDream-I1 is constructed with a new sparse Diffusion Transformer (DiT) structure. Specifically, it starts with a dual-stream decoupled design of sparse DiT with dynamic Mixture-of-Experts (MoE) architecture, in which two separate encoders are first involved to independently process image and text tokens. Then, a single-stream sparse DiT structure with dynamic MoE architecture is adopted to trigger multi-model interaction for image generation in a cost-efficient manner. To support flexiable accessibility with varied model capabilities, we provide HiDream-I1 in three variants: HiDream-I1-Full, HiDream-I1-Dev, and HiDream-I1-Fast. Furthermore, we go beyond the typical text-to-image generation and remould HiDream-I1 with additional image conditions to perform precise instruction-based editing on given images, yielding a new image editing model namely HiDream-E1. We have open-sourced all the codes and model weights of HiDream-I1 and HiDream-E1: https://github.com/HiDream-ai/HiDream-I1 and https://github.com/HiDream-ai/HiDream-E1. These models quickly gained strong traction in the community, ranking among the top globally on the Hugging Face Models Trending list within just one week of launch. In under a month, it surpassed 280,000 downloads and has been officially integrated into the Diffusers library. It is now widely adopted by leading community tools and products, including ComfyUI, Recraft, WaveSpeedAI, fal.ai, and Pruna AI - reflecting the model's growing impact across the open-source AI ecosystem.

Recent advances in diffusion-based image editing models have demonstrated remarkable success. However, these models primarily rely on high-quality textual prompts to guide image manipulation, creating a significant barrier for non-expert users. In this demonstration, we present an exemplar-based image editing framework named Edit-by-Example, which eliminates the reliance on textual prompts, requires only a single pair of before-and-after images to encapsulate the desired editing effect that can readily be applied on the user-provided query image without any model fine-tuning. Technically, our framework comprises two components: an Adaptive Editing Policy Module (AEPM) and a Generation Module (GM). The AEPM jointly analyzes cross-image relationships in exemplar pairs and query image content to derive optimal editing directions, while GM executes these policies through an off-the-shelf image editor with optional semantic alignment verification. We introduce EEdBench, a comprehensive benchmark for exemplar-based image editing containing 1,500 test cases across 15 categories. Experiments demonstrate that our framework outperforms existing prompt-free methods in editing direction accuracy (S-Visual) and fidelity (FID).

This paper demonstrates a pioneering unified multimodal agent that transforms complex visual content creation into an intuitive, conversational experience, allowing users to talk, imagine, and evolve their ideas. Overcoming the limitations of fragmented multimodal technique tools, our system seamlessly integrates text-to-image generation, instruction-based image editing, text/image-to-video generation, and interactive understanding within a single AI interface. Users of all skill levels can perform sophisticated visual tasks using natural language and visual inputs. The system's architecture features a central Coordinator module processing multimodal inputs and directing tasks to Generation or Chat pathways. For Generation, a Planner utilizes our state-of-the-art specialized models in image/video generation and image editing, while the Chat function facilitates clarification and collaboration. The interactive demonstration will showcase intuitive multimodal input, seamless real-time content creation/editing, dynamic interactive understanding, and a unified workflow. This agent pioneers a new way for accessible, interactive visual storytelling and collaborative content creation in multimodal generative AI.

Recent advancements in multimodal AIGC have enabled impressive text-to-video synthesis, but a critical challenge remains: maintaining consistent identity of key subjects across generated frames. To address this limitation, we introduce the Identity-Preserving Video Generation (IPVG) grand challenge. This challenge aims to propel the field toward more controllable generative models by focusing community efforts on preserving identity during the video generation process. To support these efforts, we publicly release the Identity-Preserving Video Benchmark (VIP-200K), a novel dataset comprising approximately 500,000 video-prompt pairs with 200,000 unique identities, each coupled with a reference identity image. Through this grand challenge and dataset, we provide a fertile ground for developing solutions that lead to more user-steerable video synthesis systems. The challenge homepage is https://hidream-ai.github.io/ipvg-challenge.github.io/.

Diffusion models have yielded remarkable success in virtual try-on (VTON) task, yet they often fall short of fully meeting user expectations regarding visual quality and detail preservation. To alleviate this issue, we curate a dataset of synthesized VTON images annotated with human judgments across multiple perceptual criteria. A vision large language model (VLLM), namely VTON-VLLM, is then learnt on these annotations. VTON-VLLM functions as a unified ``fashion expert'' and is capable of both evaluating and steering VTON synthesis towards human preferences. Technically, beyond serving as an automatic VTON evaluator, VTON-VLLM upgrades VTON model through two pivotal ways: (1) providing fine-grained supervisory signals during the training of a plug-and-play VTON refinement model, and (2) enabling adaptive and preference-aware test-time scaling at inference. To benchmark VTON models more holistically, we introduce VITON-Bench, a challenging test suite of complex try-on scenarios, and human-preference–aware metrics. Extensive experiments demonstrate that powering VTON models with our VTON-VLLM markedly enhances alignment with human preferences. Code is publicly available at: [https://github.com/HiDream-ai/VTON-VLLM/](https://github.com/HiDream-ai/VTON-VLLM/).

It is always well believed that pre-trained vision-language foundation models (e.g., CLIP) would substantially facilitate vision-language tasks. Nevertheless, there has been less evidence in support of the idea on describing novel objects in images. In this paper, we propose the Novel Object Transformer with CLIP (NOTC), a Transformer-based model that innovatively exploits the powerful vision-language representation ability of CLIP to enhance novel object captioning model’s training and sentence decoding processes. Technically, given the primary bag-of-objects extracted by Faster R-CNN, NOTC first capitalize on an object distiller module to emphasize the most salient objects and infer the missing novel ones. The refined object words are additionally fed into the object-centric word predictor to generate sentence word-by-word. During training, we design a CLIP-based self-critical sequence training paradigm to select visually-grounded sampled sentence with higher CLIP score reward, which enables a joint training process of captioning model over out-domain training images with novel objects. Moreover, at inference, a new CLIP beam search algorithm is devised to enforce the existence of novel objects and encourage the partial word sequences with higher CLIP scores, thereby decoding both visually-grounded and comprehensive sentences. Extensive experiments are conducted on held-out COCO and nocaps datasets, and competitive performances are reported when compared to state-of-the-art approaches.

Recently Vision Transformer (ViT) and Convolution Neural Network (CNN) start to emerge as a hybrid deep architecture with better model capacity, generalization, and latency trade-off. Most of these hybrid architectures often directly stack self-attention module with static convolution or fuse their outputs through two pathways within each block. Instead, we present a new Transformer architecture (namely Stream-ViT) to novelly integrate ViT with streamlined convolutions, i.e., a series of high-to-low resolution convolutions. The kernels of each convolution are dynamically learnt on a basis of current input features plus pre-learnt kernels throughout the whole network. The new architecture incorporates a critical pathway to streamline kernel generation that triggers the interactions between dynamically learnt convolutions across different layers. Moreover, the introduction of a layer-wise streamlined convolution is functionally equivalent to a squeezed version of multi-branch convolution structure, thereby improving the capacity of self-attention module with enlarged cardinality in a cost-efficient manner. We validate the superiority of Stream-ViT over multiple vision tasks, and its performances surpass state-of-the-art ViT and CNN backbones with comparable FLOPs.

Masked image modeling (MIM) has been considered as the state-of-the-art (SOTA) self-supervised learning (SSL) technique in terms of visual pretraining. The impressive generalization ability of MIM also paves the way for the remarkable success of large-scale vision foundation models. In this article, we further discuss the validity and advantages of implementing MIM techniques in the reproducing kernel Hilbert spaces (RKHSs) and we associate the analysis with a novel MIM method named R-MIM (short for RKHS-MIM). Through the careful construction of an augmentation graph and by using spectral decomposition techniques, we establish a systematic theoretical understanding between the proposed R-MIM's generalization ability and the choice of kernel function used during training. Specifically, we reach a conclusion that both of the local Lipschitz constant of the resultant R-MIM model and the corresponding expected pretraining error can have a strong composite effect on bounding downstream task error, depending on the kernel options. We demonstrate that under mild mathematical assumptions, R-MIM method is guaranteed to return a lower bound on downstream tasks in comparison to vanilla MIM techniques, such as masked autoencoder (MAE) and SimMIM. Empirical justification well corroborates our theoretical hypothesis and analysis in showing the superior generalization of the proposed R-MIM and the theoretical link to kernel choices. The code is available at: https://github.com/yurui-q/R-MIM.

Masked autoencoder (MAE) has been regarded as a capable self-supervised learner for various downstream tasks. Nevertheless, the model still lacks high-level discriminability, which results in poor linear probing performance. In view of the fact that strong augmentation plays an essential role in contrastive learning, can we capitalize on strong augmentation in MAE? The difficulty originates from the pixel uncertainty caused by strong augmentation that may affect the reconstruction, and thus, directly introducing strong augmentation into MAE often hurts the performance. In this article, we delve into the potential of strong augmented views to enhance MAE while maintaining MAE's advantages. To this end, we propose a simple yet effective masked Siamese autoencoder (MSA) model, which consists of a student branch and a teacher branch. The student branch derives MAE's advanced architecture, and the teacher branch treats the unmasked strong view as an exemplary teacher to impose high-level discrimination onto the student branch. We demonstrate that our MSA can improve the model's spatial perception capability and, therefore, globally favors interimage discrimination. Empirical evidence shows that the model pretrained by MSA provides superior performances across different downstream tasks. Notably, linear probing performance on frozen features extracted from MSA leads to 6.1% gains over MAE on ImageNet-1k. Fine-tuning (FT) the network on VQAv2 task finally achieves 67.4% accuracy, outperforming 1.6% of the supervised method DeiT and 1.2% of MAE. Codes and models are available at https://github.com/KimSoybean/MSA.

Jingtao Zhan, Qingyao Ai, Yiqun Liu, Yingwei Pan, Ting Yao, Jiaxin Mao, Shaoping Ma, Tao Mei. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Diffusion models are just at a tipping point for image super-resolution task. Nevertheless, it is not trivial to capitalize on diffusion models for video super-resolution which necessitates not only the preservation of visual appearance from low-resolution to high-resolution videos, but also the temporal consistency across video frames. In this paper, we propose a novel approach, pursuing Spatial Adaptation and Temporal Coherence (SATeCo), for video super-resolution. SATeCo pivots on learning spatial-temporal guidance from low-resolution videos to calibrate both latent-space high-resolution video denoising and pixel-space video reconstruction. Technically, SATeCo freezes all the parameters of the pre-trained UNet and VAE, and only optimizes two deliberately-designed spatial feature adaptation (SFA) and temporal feature alignment (TFA) modules, in the decoder of UNet and VAE. SFA modulates frame features via adaptively estimating affine parameters for each pixel, guaran-teeing pixel-wise guidance for high-resolution frame syn-thesis. TFA delves into feature interaction within a 3D local window (tube let) through self-attention, and executes cross-attention between tubelet and its low-resolution counterpart to guide temporal feature alignment. Extensive experiments conducted on the REDS4 and Vid4 datasets demonstrate the effectiveness of our approach.

Recent innovations on text-to-3D generation have featured Score Distillation Sampling (SDS), which enables the zero-shot learning of implicit 3D models (NeRF) by directly distilling prior knowledge from 2D diffusion models. However, current SDS-based models still struggle with intricate text prompts and commonly result in distorted 3D models with unrealistic textures or cross-view inconsistency issues. In this work, we introduce a novel Visual Prompt-guided text-to-3D diffusion model (VP3D) that explicitly unleashes the visual appearance knowledge in 2D visual prompt to boost text-to-3D generation. Instead of solely supervising SDS with text prompt, VP3D first capitalizes on 2D diffusion model to generate a high-quality image from input text, which subsequently acts as visual prompt to strengthen SDS optimization with explicit visual appearance. Mean-while, we couple the SDS optimization with additional differentiable reward function that encourages rendering images of 3D models to better visually align with 2D visual prompt and semantically match with text prompt. Through extensive experiments, we show that the 2D Visual Prompt in our VP3D significantly eases the learning of visual appearance of 3D models and thus leads to higher visual fidelity with more detailed textures. It is also appealing in view that when replacing the self-generating visual prompt with a given reference image, VP3D is able to trigger a new task of stylized text-to-3D generation. Our project page is available at https://vp3d-cvpr24.github.io.

Diffusion models have recently brought a powerful rev-olution in image generation. Despite showing impressive generative capabilities, most of these models rely on the current sample to denoise the next one, possibly resulting in denoising instability. In this paper, we reinterpret the iterative denoising process as model optimization and leverage a moving average mechanism to ensemble all the prior samples. Instead of simply applying moving average to the denoised samples at different timesteps, we first map the denoised samples to data space and then perform moving average to avoid distribution shift across timesteps. In view that diffusion models evolve the recovery from low-frequency components to high-frequency details, we fur-ther decompose the samples into different frequency components and execute moving average separately on each component. We name the complete approach “Moving Aver-age Sampling in Frequency domain (MASF)”. MASF could be seamlessly integrated into mainstream pre-trained dif-fusion models and sampling schedules. Extensive experi-ments on both unconditional and conditional diffusion mod-els demonstrate that our MASF leads to superior performances compared to the baselines, with almost negligible additional complexity cost.

Recent advances in text-to- video generation have demonstrated the utility of powerful diffusion models. Nev-ertheless, the problem is not trivial when shaping diffusion models to animate static image (i.e., image-to-video generation). The difficulty originates from the aspect that the diffusion process of subsequent animated frames should not only preserve the faithful alignment with the given image but also pursue temporal coherence among adjacent frames. To alleviate this, we present TRIP, a new recipe of image-to-video diffusion paradigm that pivots on image noise prior derived from static image to jointly trigger inter-frame relational reasoning and ease the coherent temporal modeling via temporal residual learning. Technically, the image noise prior is first attained through one-step backward dif-fusion process based on both static image and noised video latent codes. Next, TRIP executes a residual-like dual-path scheme for noise prediction: 1) a shortcut path that directly takes image noise prior as the reference noise of each frame to amplify the alignment between the first frame and sub-sequent frames; 2) a residual path that employs 3D-UNet over noised video and static image latent codes to enable inter-frame relational reasoning, thereby easing the learning of the residual noise for each frame. Furthermore, both reference and residual noise of each frame are dynamically merged via attention mechanism for final video generation. Extensive experiments on WebVid-10M, DTDB and MSR-VTT datasets demonstrate the effectiveness of our TRIP for image-to-video generation. Please see our project page at https://trip-i2v.github.io/TRIP/.

Diffusion Transformer (DiT) has emerged as the new trend of generative diffusion models on image generation. In view of extremely slow convergence in typical DiT, recent breakthroughs have been driven by mask strategy that significantly improves the training efficiency of DiT with additional intra-image contextual learning. Despite this progress, mask strategy still suffers from two inherent limitations: (a) training-inference discrepancy and (b) fuzzy relations between mask reconstruction & generative diffusion process, resulting in sub-optimal training of DiT. In this work, we address these limitations by novelly unleashing the self-supervised discrimination knowledge to boost DiT training. Technically, we frame our DiT in a teacher-student manner. The teacher-student discriminative pairs are built on the diffusion noises along the same Probability Flow Ordinary Differential Equation (PF-ODE). Instead of applying mask reconstruction loss over both DiT encoder and decoder, we decouple DiT encoder and decoder to separately tackle discriminative and generative objectives. In particular, by encoding discriminative pairs with student and teacher DiT encoders, a new discriminative loss is designed to encourage the inter-image alignment in the selfsupervised embedding space. After that, student samples are fed into student DiT decoder to perform the typical generative diffusion task. Extensive experiments are conducted on ImageNet dataset, and our method achieves a competitive balance between training cost and generative capacity.

Despite having tremendous progress in image-to-3D generation, existing methods still struggle to produce multi-view consistent images with high-resolution textures in detail, especially in the paradigm of 2D diffusion that lacks 3D awareness. In this work, we present High-resolution Image-to-3D model (Hi3D), a new video diffusion based paradigm that redefines a single image to multi-view images as 3D-aware sequential image generation (i.e., orbital video generation). This methodology delves into the underlying temporal consistency knowledge in video diffusion model that generalizes well to geometry consistency across multiple views in 3D generation. Technically, Hi3D first empowers the pre-trained video diffusion model with 3D-aware prior (camera pose condition), yielding multi-view images with low-resolution texture details. A 3D-aware video-to-video refiner is learnt to further scale up the multi-view images with high-resolution texture details. Such high-resolution multi-view images are further augmented with novel views through 3D Gaussian Splatting, which are finally leveraged to obtain high-fidelity meshes via 3D reconstruction. Extensive experiments on both novel view synthesis and single view reconstruction demonstrate that our Hi3D manages to produce superior multi-view consistency images with highly-detailed textures. Source code and data are available at https://github.com/yanghb22-fdu/Hi3D-Official.

Text-driven human motion generation, which creates motion sequences based on textual descriptions, has attracted great attention in the communities of multimedia and artificial intelligence. By parsing and comprehending textual information and converting it into specific human movements, it realizes a direct transformation from human semantics to motion sequences. New text-driven human motion generators are springing up to achieve better performance. However, the absence of well-trained evaluators that can effectively estimate the consistency between the text prompts and motions generated by existing generators remains a challenge. To address the above issues, we propose an open-source library with a powerful Contrastive Language-and-Motion (CLaM) pre-training evaluator, which can be employed for evaluating a variety of text-driven human motion generation algorithms. We perform a thorough performance evaluation of the existing algorithms on various metrics, such as R-Precision. As a by-product, we build a large-scale HumanML3D-synthesis dataset, which consists of 14,616 motion sequences and 547,102 textual descriptions, which is ten times larger than the widely-used HumanML3D dataset. The source codes and models for CLaM are available at~https://github.com/SheldongChen/CLaM/.

The emergence of text-to-image generation models has led to the recognition that image enhancement, performed as post-processing, would significantly improve the visual quality of the generated images. Exploring diffusion models to enhance the generated images nevertheless is not trivial and necessitates to delicately enrich plentiful details while preserving the visual appearance of key content in the original image. In this paper, we propose a novel framework, namely FreeEnhance, for content-consistent image enhancement using the off-the-shelf image diffusion models. Technically, FreeEnhance is a two-stage process that firstly adds random noise to the input image and then capitalizes on a pre-trained image diffusion model (i.e., Latent Diffusion Models) to denoise and enhance the image details. In the noising stage, FreeEnhance is devised to add lighter noise to the region with higher frequency to preserve the high-frequent patterns (e.g., edge, corner) in the original image. In the denoising stage, we present three target properties as constraints to regularize the predicted noise, enhancing images with high acutance and high visual quality. Extensive experiments conducted on the HPDv2 dataset demonstrate that our FreeEnhance outperforms the state-of-the-art image enhancement models in terms of quantitative metrics and human preference. More remarkably, FreeEnhance also shows higher human preference compared to the commercial image enhancement solution of Magnific AI.

The hybrid deep models of Vision Transformer (ViT) and Convolution Neural Network (CNN) have emerged as a powerful class of backbones for vision tasks. Scaling up the input resolution of such hybrid backbones naturally strengthes model capacity, but inevitably suffers from heavy computational cost that scales quadratically. Instead, we present a new hybrid backbone with HIgh-Resolution Inputs (namely HIRI-ViT), that upgrades prevalent four-stage ViT to five-stage ViT tailored for high-resolution inputs. HIRI-ViT is built upon the seminal idea of decomposing the typical CNN operations into two parallel CNN branches in a cost-efficient manner. One high-resolution branch directly takes primary high-resolution features as inputs, but uses less convolution operations. The other low-resolution branch first performs down-sampling and then utilizes more convolution operations over such low-resolution features. Experiments on both recognition task (ImageNet-1K dataset) and dense prediction tasks (COCO and ADE20 K datasets) demonstrate the superiority of HIRI-ViT. More remarkably, under comparable computational cost ( ∼ 5.0 GFLOPs), HIRI-ViT achieves to-date the best published Top-1 accuracy of 84.3% on ImageNet with 448×448 inputs, which absolutely improves 83.4% of iFormer-S by 0.9% with 224×224 inputs.

By formulating the data generation as a sequence procedure of denoising autoencoding, diffusion models have achieved superior in-painting performance on image data and beyond. Nevertheless, it is not trivial when capitalizing on diffusion models to generate missing 3D points. The difficulty originates from the intrinsic structure where 3D point cloud is a set of unordered and irregular coordinates. That motivates us to delve into the 3D structural information for designing point cloud encoder-decoder and shape latent generator, to precisely formulate the latent distribution of the complete point cloud and partial observation. In this paper, we propose Point cloud completion with Latent Diffusion Models (PointLDM), a new approach that leverages the conditional denoising diffusion probabilistic modeling (DDPM) in the 3D latent space for shape reconstruction. The architecture of PointLDM consists of a transformer-based variational auto-encoder (VAE) to model the complete shape latent, and a diffusion network for shape latent prediction. The encoder of VAE exploits both of global shape latent and local point features in shape distribution learning. With the learnt shape latent, the decoder first decodes the shape latent into coarse points, and then recovers the fine-grained details around each coarse point by deforming a 2D grid. To reconstruct the shape latent from partial observation, the diffusion network treats the partial observation as the conditional input and generates the shape latent via DDPM. Extensive experiments conducted on MVP, Completion3D, and KITTI quantitatively and qualitatively demonstrate the efficacy of PointLDM over the state-of-the-art shape completion approaches.

Point cloud analysis faces computational system overhead, limiting its application on mobile or edge devices. Directly employing small models may result in a significant drop in performance since it is difficult for a small model to adequately capture local structure and global shape information simultaneously, which are essential clues for point cloud analysis. This paper explores feature distillation for lightweight point cloud models. To mitigate the semantic gap between the lightweight student and the cumbersome teacher, we propose bidirectional knowledge reconfiguration (BKR) to distill informative contextual knowledge from the teacher to the student. Specifically, a top-down knowledge reconfiguration and a bottom-up knowledge reconfiguration are developed to inherit diverse local structure information and consistent global shape knowledge from the teacher, respectively. However, due to the farthest point sampling in most point cloud models, the intermediate features between teacher and student are misaligned, deteriorating the feature distillation performance. To eliminate it, we propose a feature mover's distance (FMD) loss based on optimal transportation, which can measure the distance between unordered point cloud features effectively. Extensive experiments conducted on shape classification, part segmentation, and semantic segmentation benchmarks demonstrate the universality and superiority of our method.

Temporal dynamics represent the evolving of video content over time, which are critical for action recognition. In this paper, we ask the question: can the off-the-shelf image transformer architecture learn temporal dynamics in videos? To this end, we propose Multidimensional Stacked Image (MSImage) as a new arrangement of video data, which can be fed to image transformers. Technically, MSImage is a high-resolution image that is composed of several evenly-sampled video clips stacked along the channel and space dimensions. The frames in each clip are concatenated along the channel dimension for the transformers to infer short-term dynamics. Meanwhile, the clips are arranged on different spatial positions for learning long-term dynamics. On this basis we propose MSImageFormer-a new variant of image transformer that takes MSImage as the input and is jointly optimized by video classification loss and a new dynamics enhancement loss. The network optimization attends to the high-frequency component of MSImage, avoiding overfitting to static visual patterns. We empirically demonstrate the merits of the MSImageFormer on six action recognition benchmarks. With only 2D image transformer as the classifier, our MSImageFormer achieves 85.3% and 69.7% top-1 accuracy on Kinetics-400 and Something-Something V2 datasets, respectively. Despite with fewer computations, the results are comparable to the SOTA 3D CNNs and video transformers.

Learning proper representations for speech and gesture is essential for co-speech gesture generation. Existing approaches either utilize direct representations or independently encode the speech and gesture, which neglect the joint representation to highlight the interplay between these two modalities. In this work, we propose a novel Cross-modal Quantization (CMQ) to jointly learn the quantized codes for speech and gesture together. Such representation highlights the speech-gesture interaction before actually learning the complex mapping, and thus better suits the intricate mapping between speech and gesture. Specifically, the Cross-modal Quantizer jointly encodes speech and gesture as discrete codebooks, enabling better cross-modal interaction. Cross-modal Predictor subsequently utilizes the learned codebooks to autoregressively predict the next-step gesture. With cross-modal quantization, our approach yields much higher codebook usage and generates more realistic and diverse gestures in practice. Extensive experiments are conducted on both 3D and 2D datasets as well as the subjective user study, demonstrating a clear performance gain compared to several baseline models in terms of audio-visual alignment and gesture diversity. In particular, our method demonstrates a three-fold improvement in diversity compared to baseline models, while simultaneously maintaining high motion fidelity.

Video scene graph generation has been an emerging research topic, which aims to interpret a video as a temporally-evolving graph structure by representing video objects as nodes and their relations as edges. Existing approaches predominantly follow a multi-step scheme, including frame-level object detection, relation recognition and temporal association. Although effective, these approaches neglect the mutual interactions between independent steps, resulting in a sub-optimal solution. We present a novel end-to-end framework for video scene graph generation, which naturally unifies object detection, object tracking, and relation recognition via a new Transformer structure, namely Temporal Propagation Transformer (TPT). Particularly, TPT extends the existing Transformer-based object detector (e.g., DETR) along the temporal dimension by involving a query propagation module, which can additionally associate the detected instances by identities across frames. A temporal dynamics encoder is then leveraged to dynamically enrich the features of the detected instances for relation recognition by attending to their historic states in previous frames. Meanwhile, the relation propagation strategy is devised to emphasize the temporal consistency of relation recognition results among adjacent frames. Extensive experiments conducted on VidHOI and Action Genome benchmarks demonstrate the superior performance of the proposed TPT over the state-of-the-art methods.

Vision transformer (ViT) has become a leading tool in various computer vision tasks, owing to its unique self-attention mechanism that learns visual representations explicitly through cross-patch information interactions. Despite having good success, the literature seldom explores the explainability of ViT, and there is no clear picture of how the attention mechanism with respect to the correlation across comprehensive patches will impact the performance and what is the further potential. In this work, we propose a novel explainable visualization approach to analyze and interpret the crucial attention interactions among patches for ViT. Specifically, we first introduce a quantification indicator to measure the impact of patch interaction and verify such quantification on attention window design and indiscriminative patches removal. Then, we exploit the effective responsive field of each patch in ViT and devise a window-free transformer (WinfT) architecture accordingly. Extensive experiments on ImageNet demonstrate that the exquisitely designed quantitative method is shown able to facilitate ViT model learning, leading the top-1 accuracy by 4.28% at most. More remarkably, the results on downstream fine-grained recognition tasks further validate the generalization of our proposal.

Some cognitive research has discovered that humans accomplish event segmentation as a side effect of event anticipation. Inspired by this discovery, we propose a simple yet effective end-to-end self-supervised learning framework for event segmentation/boundary detection. Unlike the mainstream clustering-based methods, our framework exploits a transformer-based feature reconstruction scheme to detect event boundaries by reconstruction errors. This is consistent with the fact that humans spot new events by leveraging the deviation between their prediction and what is perceived. Thanks to their heterogeneity in semantics, the frames at boundaries are difficult to be reconstructed (generally with large reconstruction errors), which is favorable for event boundary detection. In addition, since the reconstruction occurs on the semantic feature level instead of the pixel level, we develop a temporal contrastive feature embedding (TCFE) module to learn the semantic visual representation for frame feature reconstruction (FFR). This procedure is like humans building up experiences with "long-term memory." The goal of our work is to segment generic events rather than localize some specific ones. We focus on achieving accurate event boundaries. As a result, we adopt the F1 score (Precision/Recall) as our primary evaluation metric for a fair comparison with previous approaches. Meanwhile, we also calculate the conventional frame-based mean over frames (MoF) and intersection over union (IoU) metric. We thoroughly benchmark our work on four publicly available datasets and demonstrate much better results. The source code is available at https://github.com/wang3702/CoSeg.

The recent advances in deep learning predominantly construct models in their internal representations, and it is opaque to explain the rationale behind and decisions to human users. Such explainability is especially essential for domain adaptation, whose challenges require developing more adaptive models across different domains. In this article, we ask the question: How much does each sample in the source domain contribute to the network’s prediction on the samples from the target domain? To address this, we devise a novel Interpretable Deep Classifier (IDC) that learns the nearest source samples of a target sample as evidence upon which the classifier makes the decision. Technically, IDC maintains a differentiable memory bank for each category, and the memory slot derives a form of key–value pair. The key records the features of discriminative source samples, and the value stores the corresponding properties, e.g., representative scores of the features for describing the category. IDC computes the loss between the output of IDC and the labels of source samples to back-propagate to adjust the representative scores and update the memory banks. Extensive experiments on Office-Home and VisDA-2017 datasets demonstrate that our IDC leads to a more explainable model with almost no accuracy degradation and effectively calibrates classification for optimum reject options. More remarkably, when taking IDC as a prior interpreter, capitalizing on 0.1% source training data selected by IDC still yields superior results than that uses full training set on VisDA-2017 for unsupervised domain adaptation.

Scene graph generation (SGG) aims to abstract an image into a graph structure, by representing objects as graph nodes and their relations as labeled edges. However, two knotty obstacles limit the practicability of current SGG methods in real-world scenarios: 1) training SGG models requires time-consuming ground-truth annotations, and 2) the closed-set object categories make the SGG models limited in their ability to recognize novel objects outside of training corpora. To address these issues, we novelly exploit a powerful pre-trained visual-semantic space (VSS) to trigger language-supervised and open-vocabulary SGG in a simple yet effective manner. Specifically, cheap scene graph supervision data can be easily obtained by parsing image language descriptions into semantic graphs. Next, the noun phrases on such semantic graphs are directly grounded over image regions through region-word alignment in the pre-trained VSS. In this way, we enable open-vocabulary object detection by performing object category name grounding with a text prompt in this VSS. On the basis of visually-grounded objects, the relation representations are naturally built for relation recognition, pursuing open-vocabulary SGG. We validate our proposed approach with extensive experiments on the Visual Genome benchmark across various SGG scenarios (i.e., supervised / language-supervised, closed-set / open-vocabulary). Consistent superior performances are achieved compared with existing methods, demonstrating the potential of exploiting pre-trained VSS for SGG in more practical scenarios.

Point cloud completion aims to recover the completed 3D shape of an object from its partial observation. A common strategy is to encode the observed points to a global feature vector and then predict the complete points through a generative process on this vector. Nevertheless, the results may suffer from the high-quality shape generation problem due to the fact that a global feature vector cannot sufficiently characterize diverse patterns in one object. In this paper, we present a new shape completion architecture, namely AnchorFormer, that innovatively leverages pattern-aware discriminative nodes, i.e., anchors, to dynamically capture regional information of objects. Technically, AnchorFormer models the regional discrimination by learning a set of anchors based on the point features of the input partial observation. Such anchors are scattered to both observed and unobserved locations through estimating particular offsets, and form sparse points together with the down-sampled points of the input observation. To reconstruct the finegrained object patterns, AnchorFormer further employs a modulation scheme to morph a canonical 2D grid at individual locations of the sparse points into a detailed 3D structure. Extensive experiments on the PCN, ShapeNet-55/34 and KITTI datasets quantitatively and qualitatively demonstrate the efficacy of AnchorFormer over the state-of-the-art point cloud completion approaches. Source code is available at https://github.com/chenzhik/AnchorFormer.

Feature invariance under different data transformations, i.e., transformation invariance, can be regarded as a type of self-supervision for representation learning. In this paper, we present PointClustering, a new unsupervised representation learning scheme that leverages transformation invariance for point cloud pre-training. PointClustering formulates the pretext task as deep clustering and employs transformation invariance as an inductive bias, following the philosophy that common point cloud transformation will not change the geometric properties and semantics. Technically, PointClustering iteratively optimizes the feature clusters and backbone, and delves into the transformation invariance as learning regularization from two perspectives: point level and instance level. Point-level invariance learning maintains local geometric properties through gathering point features of one instance across transformations, while instance-level invariance learning further measures clusters over the entire dataset to explore semantics of instances. Our PointClustering is architecture-agnostic and readily applicable to MLP-based, CNN-based and Transformer-based backbones. We empirically demonstrate that the models pre-learnt on the ScanNet dataset by PointClustering provide superior performances on six benchmark-s, across downstream tasks of classification and segmentation. More remarkably, PoinrClustering achieves an accuracy of 94.5% on ModelNet40 with Transformer backbone. Source code is available at https://github.com/FuchenUSTC/PointClustering.

Recent advances on text-to-image generation have witnessed the rise of diffusion models which act as powerful generative models. Nevertheless, it is not trivial to exploit such latent variable models to capture the dependency among discrete words and meanwhile pursue complex visual-language alignment in image captioning. In this paper, we break the deeply rooted conventions in learning Transformer-based encoder-decoder, and propose a new diffusion model based paradigm tailored for image captioning, namely Semantic-Conditional Diffusion Networks (SCD-Net). Technically, for each input image, we first search the semantically relevant sentences via cross-modal retrieval model to convey the comprehensive semantic information. The rich semantics are further regarded as semantic prior to trigger the learning of Diffusion Transformer, which produces the output sentence in a diffusion process. In SCD-Net, multiple Diffusion Transformer structures are stacked to progressively strengthen the output sentence with better visional-language alignment and linguistical coherence in a cascaded manner. Furthermore, to stabilize the diffusion process, a new self-critical sequence training strategy is designed to guide the learning of SCD-Net with the knowledge of a standard autoregressive Transformer model. Extensive experiments on COCO dataset demonstrate the promising potential of using diffusion models in the challenging image captioning task. Source code is available at

Deep neural networks (DNNs) usually fail to generalize well to outside of distribution (OOD) data, especially in the extreme case of single domain generalization (single-DG) that transfers DNNs from single domain to multiple unseen domains. Existing single-DG techniques commonly devise various data-augmentation algorithms, and remould the multi-source domain generalization methodology to learn domain-generalized (semantic) features. Nevertheless, these methods are typically modality-specific, thereby being only applicable to one single modality (e.g., image). In contrast, we target a versatile Modality-Agnostic Debiasing (MAD) framework for single-DG, that enables generalization for different modalities. Technically, MAD introduces a novel two-branch classifier: a biased-branch encourages the classifier to identify the domain-specific (superficial) features, and a general-branch captures domain-generalized features based on the knowledge from biased-branch. Our MAD is appealing in view that it is pluggable to most single-DG models. We validate the superiority of our MAD in a variety of single-DG scenarios with different modalities, including recognition on 1D texts, 2D images, 3D point clouds, and semantic segmentation on 2D images. More remarkably, for recognition on 3D point clouds and semantic segmentation on 2D images, MAD improves DSU by 2.82% and 1.5% in accuracy and mIOU.

Although the estimation of 3D human pose and shape (HPS) is rapidly progressing, current methods still cannot reliably estimate moving humans in global coordinates, which is critical for many applications. This is particularly challenging when the camera is also moving, entangling human and camera motion. To address these issues, we adopt a novel 5D representation (space, time, and identity) that enables end-to-end reasoning about people in scenes. Our method, called TRACE, introduces several novel architectural components. Most importantly, it uses two new “maps” to reason about the 3D trajectory of people over time in camera, and world, coordinates. An additional memory unit enables persistent tracking of people even during long occlusions. TRACE is the first one-stage method to jointly recover and track 3D humans in global coordinates from dynamic cameras. By training it end-to-end, and using full image information, TRACE achieves state-of-the-art performance on tracking and HPS benchmarks. The code11https://www.yusun.work/TRACE/TRACE.html and dataset22https://github.com/Arthur151/DynaCam are released for research purposes.

Parsing an unstructured point set into constituent local geometry structures (e.g., edges or surfaces) would be helpful for understanding and representing point clouds. This motivates us to design a deep architecture to model the hierarchical geometry from points, edges, surfaces (triangles), to super-surfaces (adjacent surfaces) for the thorough analysis of point clouds. In this paper, we present a novel Hierarchical Geometry Network (HGNet) that integrates such hierarchical geometry structures from super-surfaces, surfaces, edges, to points in a top-down manner for learning point cloud representations. Technically, we first construct the edges between every two neighbor points. A point-level representation is learnt with edge-to-point aggregation, i.e., aggregating all connected edges into the anchor point. Next, as every two neighbor edges compose a surface, we obtain the edge-level representation of each anchor edge via surface-to-edge aggregation over all neighbor surfaces. Furthermore, the surface-level representation is achieved through super-surface-to-surface aggregation by transforming all super-surfaces into the anchor surface. A Transformer structure is finally devised to unify all the point-level, edge-level, and surface-level features into the holistic point cloud representations. Extensive experiments on four point cloud analysis datasets demonstrate the superiority of HGNet for 3D object classification and part/semantic segmentation tasks. More remarkably, HGNet achieves the overall accuracy of 89.2% on ScanObjectNN, improving PointNeXt-S by 1.5%.

Dynamically synthesizing talking speech that actively responds to a listening head is critical during the face-to-face interaction. For example, the speaker could take advantage of the listener’s facial expression to adjust the tones, stressed syllables, or pauses. In this work, we present a new visual-aware text-to-speech (VA-TTS) task to synthesize speech conditioned on both textual inputs and sequential visual feedback (e.g., nod, smile) of the listener in face-to-face communication. Different from traditional text-to-speech, VA-TTS highlights the impact of visual modality. On this newly-minted task, we devise a baseline model to fuse phoneme linguistic information and listener visual signals for speech synthesis. Extensive experiments on multimodal conversation dataset ViCo-X verify our proposal for generating more natural audio with scenario-appropriate rhythm and prosody.

Recent progress on multi-modal 3D object detection has featured BEV (Bird-Eye-View) based fusion, which effectively unifies both LiDAR point clouds and camera images in a shared BEV space. Nevertheless, it is not trivial to perform camera-to-BEV transformation due to the inherently ambiguous depth estimation of each pixel, resulting in spatial misalignment between these two multi-modal features. Moreover, such transformation also inevitably leads to projection distortion of camera image features in BEV space. In this paper, we propose a novel Object-centric Fusion (ObjectFusion) paradigm, which completely gets rid of camera-to-BEV transformation during fusion to align object-centric features across different modalities for 3D object detection. ObjectFusion first learns three kinds of modality-specific feature maps (i.e., voxel, BEV, and image features) from LiDAR point clouds and its BEV projections, camera images. Then a set of 3D object proposals are produced from the BEV features via a heatmap-based proposal generator. Next, the 3D object proposals are reprojected back to voxel, BEV, and image spaces. We leverage voxel and RoI pooling to generate spatially aligned object-centric features for each modality. All the object-centric features of three modalities are further fused at object level, which is finally fed into the detection heads. Extensive experiments on nuScenes dataset demonstrate the superiority of our ObjectFusion, by achieving 69.8% mAP on nuScenes validation set and improving BEVFusion by 1.3%.

Recent progress on multi-view 3D object reconstruction has featured neural implicit surfaces via learning high-fidelity radiance fields. However, most approaches hinge on the visual hull derived from cost-expensive silhouette masks to obtain object surfaces. In this paper, we propose a novel Object-aware Radiance Fields (ORF) to automatically learn an object-aware geometry reconstruction. The geometric correspondences between multi-view 2D object regions and 3D implicit/explicit object surfaces are additionally exploited to boost the learning of object surfaces. Technically, a critical transparency discriminator is designed to distinguish the object-intersected and object-bypassed rays based on the estimated 2D object regions, leading to 3D implicit object surfaces. Such implicit surfaces can be directly converted into explicit object surfaces (e.g., meshes) via marching cubes. Then, we build the geometric correspondence between 2D planes and 3D meshes by rasterization, and project the estimated object regions into 3D explicit object surfaces by aggregating the object information across multiple views. The aggregated object information in 3D explicit object surfaces is further reprojected back to 2D planes, aiming to update 2D object regions and enforce them to be multi-view consistent. Extensive experiments on DTU and BlendedMVS verify the capability of ORF to produce comparable surfaces against the state-of-the-art models that demand silhouette masks.

We demonstrate an automatic 3D creation system, which can create realistic 3D assets solely from a text or image prompt without requiring any specialized 3D modeling skills. Users can either describe the object they envision in natural language or upload a reference image that records what they have seen with the phone. Our system will generate a high-quality 3D mesh that faithfully matches the users' input. We propose a coarse-to-fine framework to achieve this goal. Specifically, we first obtain a low-resolution mesh instantly by utilizing a pre-trained text/image conditional 3D generative model. Using such coarse mesh as the initialization, we further optimize a high-resolution textured 3D mesh with fine-grained appearance guidance from large-scale 2D diffusion models. Our system can create visually-pleasing results in minutes, which is significantly faster than existing methods. Meanwhile, the system ensures that the resulting 3D assets are precisely aligned with the input text or image prompt. With these advanced capabilities, our demonstration provides a streamlined and intuitive platform for users to incorporate 3D creation into their daily lives.

Recent remarkable advances in large-scale text-to-image diffusion models have inspired a significant breakthrough in text-to-3D generation, pursuing 3D content creation solely from a given text prompt. However, existing text-to-3D techniques lack a crucial ability in the creative process: interactively control and shape the synthetic 3D contents according to users' desired specifications (e.g., sketch). To alleviate this issue, we present the first attempt for text-to-3D generation conditioning on the additional hand-drawn sketch, namely Control3D, which enhances controllability for users. In particular, a 2D conditioned diffusion model (ControlNet) is remoulded to guide the learning of 3D scene parameterized as NeRF, encouraging each view of 3D scene aligned with the given text prompt and hand-drawn sketch. Moreover, we exploit a pre-trained differentiable photo-to-sketch model to directly estimate the sketch of the rendered image over synthetic 3D scene. Such estimated sketch along with each sampled view is further enforced to be geometrically consistent with the given sketch, pursuing better controllable text-to-3D generation. Through extensive experiments, we demonstrate that our proposal can generate accurate and faithful 3D scenes that align closely with the input text prompts and sketches.

Recently, the multimedia community has witnessed the rise of diffusion models trained on large-scale multi-modal data for visual content creation, particularly in the field of text-to-image generation. In this paper, we propose a new task for "stylizing'' text-to-image models, namely text-driven stylized image generation, that further enhances editability in content creation. Given input text prompt and style image, this task aims to produce stylized images which are both semantically relevant to input text prompt and meanwhile aligned with the style image in style. To achieve this, we present a new diffusion model (ControlStyle) via upgrading a pre-trained text-to-image model with a trainable modulation network enabling more conditions of text prompts and style images. Moreover, diffusion style and content regularizations are simultaneously introduced to facilitate the learning of this modulation network with these diffusion priors, pursuing high-quality stylized text-to-image generation. Extensive experiments demonstrate the effectiveness of our ControlStyle in producing more visually pleasing and artistic results, surpassing a simple combination of text-to-image model and conventional style transfer techniques.

General Instance Re-identification is a very important task in computer vision, which can be widely used in many practical applications, such as person/vehicle re-identification, face recognition, wildlife protection, commodity tracing, snapshots, and so on. To meet the increasing application demand for general instance re-identification, we present FastReID as a widely used software system. In FastReID, the highly modular and extensible design makes it easy for the researcher to achieve new research ideas. Friendly manageable system configuration and engineering deployment functions allow practitioners to quickly deploy models into productions. We have implemented some state-of-the-art projects, including person re-id, partial re-id, cross-domain re-id, and vehicle re-id. Moreover, we plan to release these pre-trained models on multiple benchmark datasets. FastReID is by far the most general and high-performance toolbox that supports single and multiple GPU servers, it can reproduce our project results very easily. The source codes and models have been released at https://github.com/JDAI-CV/fast-reid.

3D content creation via text-driven stylization has played a fundamental challenge to multimedia and graphics community. Recent advances of cross-modal foundation models (e.g., CLIP) have made this problem feasible. Those approaches commonly leverage CLIP to align the holistic semantics of stylized mesh with the given text prompt. Nevertheless, it is not trivial to enable more controllable stylization of fine-grained details in 3D meshes solely based on such semantic-level cross-modal supervision. In this work, we propose a new 3DStyle-Diffusion model that triggers fine-grained stylization of 3D meshes with additional controllable appearance and geometric guidance from 2D Diffusion models. Technically, 3DStyle-Diffusion first parameterizes the texture of 3D mesh into reflectance properties and scene lighting using implicit MLP networks. Meanwhile, an accurate depth map of each sampled view is achieved conditioned on 3D mesh. Then, 3DStyle-Diffusion leverages a pre-trained controllable 2D Diffusion model to guide the learning of rendered images, encouraging the synthesized image of each view semantically aligned with text prompt and geometrically consistent with depth map. This way elegantly integrates both image rendering via implicit MLP networks and diffusion process of image synthesis in an end-to-end fashion, enabling a high-quality fine-grained stylization of 3D meshes. We also build a new dataset derived from Objaverse and the evaluation protocol for this task. Through both qualitative and quantitative experiments, we validate the capability of our 3DStyle-Diffusion. Source code and data are available at https://github.com/yanghb22-fdu/3DStyle-Diffusion-Official.

A reliable and comprehensive evaluation metric that aligns with manual preference assessments is crucial for conversational head video synthesis methods development. Existing quantitative evaluations often fail to capture the full complexity of human preference, as they only consider limited evaluation dimensions. Qualitative evaluations and user studies offer a solution but are time-consuming and labor-intensive. This limitation hinders the advancement of conversational head generation algorithms and systems. In this paper, we propose a novel learning-based evaluation metric named Preference Score (PS) for fitting human preference according to the quantitative evaluations across different dimensions. PS can serve as a quantitative evaluation without the need for human annotation. Experimental results validate the superiority of Preference Score in aligning with human perception, and also demonstrate robustness and generalizability to unseen data, making it a valuable tool for advancing conversation head generation. We expect this metric could facilitate new advances in conversational head generation. Project page: https://github.com/dc3ea9f/PreferenceScore.

Data augmentation is practically helpful for visual recognition, especially at the time of data scarcity. However, such success is only limited to quite a few light augmentations (e.g., random crop, flip). Heavy augmentations are either unstable or show adverse effects during training, owing to the big gap between the original and augmented images. This paper introduces a novel network design, noted as Augmentation Pathways (AP), to systematically stabilize training on a much wider range of augmentation policies. Notably, AP tames various heavy data augmentations and stably boosts performance without a careful selection among augmentation policies. Unlike traditional single pathway, augmented images are processed in different neural paths. The main pathway handles the light augmentations, while other pathways focus on the heavier augmentations. By interacting with multiple paths in a dependent manner, the backbone network robustly learns from shared visual patterns among augmentations, and suppresses the side effect of heavy augmentations at the same time. Furthermore, we extend AP to high-order versions for high-order scenarios, demonstrating its robustness and flexibility in practical usage. Experimental results on ImageNet demonstrate the compatibility and effectiveness on a much wider range of augmentations, while consuming fewer parameters and lower computational costs at inference time.

Transformer with self-attention has led to the revolutionizing of natural language processing field, and recently inspires the emergence of Transformer-style architecture design with competitive results in numerous computer vision tasks. Nevertheless, most of existing designs directly employ self-attention over a 2D feature map to obtain the attention matrix based on pairs of isolated queries and keys at each spatial location, but leave the rich contexts among neighbor keys under-exploited. In this work, we design a novel Transformer-style module, i.e., Contextual Transformer (CoT) block, for visual recognition. Such design fully capitalizes on the contextual information among input keys to guide the learning of dynamic attention matrix and thus strengthens the capacity of visual representation. Technically, CoT block first contextually encodes input keys via a 3×3 convolution, leading to a static contextual representation of inputs. We further concatenate the encoded keys with input queries to learn the dynamic multi-head attention matrix through two consecutive 1×1 convolutions. The learnt attention matrix is multiplied by input values to achieve the dynamic contextual representation of inputs. The fusion of the static and dynamic contextual representations are finally taken as outputs. Our CoT block is appealing in the view that it can readily replace each 3×3 convolution in ResNet architectures, yielding a Transformer-style backbone named as Contextual Transformer Networks (CoTNet). Through extensive experiments over a wide range of applications (e.g., image recognition, object detection, instance segmentation, and semantic segmentation), we validate the superiority of CoTNet as a stronger backbone. Source code is available at https://github.com/JDAI-CV/CoTNet.

Unsupervised learning is just at a tipping point where it could really take off. Among these approaches, contrastive learning has led to state-of-the-art performance. In this paper, we construct a novel probabilistic graphical model that effectively incorporates the low rank promoting prior into the framework of contrastive learning, referred to as LORAC. In contrast to the existing conventional self-supervised approaches that only considers independent learning, our hypothesis explicitly requires that all the samples belonging to the same instance class lie on the same subspace with small dimension. This heuristic poses particular joint learning constraints to reduce the degree of freedom of the problem during the search of the optimal network parameterization. Most importantly, we argue that the low rank prior employed here is not unique, and many different priors can be invoked in a similar probabilistic way, corresponding to different hypotheses about underlying truth behind the contrastive features. Empirical evidences show that the proposed algorithm clearly surpasses the state-of-the-art approaches on multiple benchmarks, including image classification, object detection, instance segmentation and keypoint detection. Code is available: https://github.com/ssl-codelab/lorac.

Recent advances have presented several strategies to mitigate the computations of self-attention mechanism with high-resolution inputs. Many of these works consider decomposing the global self-attention procedure over image patches into regional and local feature extraction procedures that each incurs a smaller computational complexity. Despite good efficiency, these approaches seldom explore the holistic interactions among all patches, and are thus difficult to fully capture the global semantics. In this paper, we propose a novel Transformer architecture that elegantly exploits the global semantics for self-attention learning, namely Dual Vision Transformer (Dual-ViT). The new architecture incorporates a critical semantic pathway that can more efficiently compress token vectors into global semantics with reduced order of complexity. Such compressed global semantics then serve as useful prior information in learning finer local pixel level details, through another constructed pixel pathway. The semantic pathway and pixel pathway are integrated together and are jointly trained, spreading the enhanced self-attention information in parallel through both of the pathways. Dual-ViT is henceforth able to capitalize on global semantics to boost self-attention learning without compromising much computational complexity. We empirically demonstrate that Dual-ViT provides superior accuracy than SOTA Transformer architectures with comparable training complexity.

Learning to reject is a special kind of self-awareness (the ability to know what you do not know), which is an essential factor for humans to become smarter. Although machine intelligence has become very accurate nowadays, it lacks such kind of self-awareness and usually acts as omniscient, resulting in overconfident errors. This article presents a comprehensive overview of this topic from three perspectives: confidence, calibration, and discrimination. Confidence is an important measurement for the reliability of model predictions. Rejection can be realized by setting thresholds on confidence. However, most models, especially modern deep neural networks, are usually overconfident. Therefore, calibration is a process to ensure confidence matching the actual likelihood of correctness, including two approaches: post-calibration and self-calibration. Calibration reflects the global characteristic of confidence, and the local distinguishing property of confidence is also important. In light of this, discrimination focuses on the performance of accepting positive samples while rejecting negative samples. As a binary classification problem, the challenge of discrimination comes from the missing and nonrepresentativeness of the negative data. Three discrimination tasks are comprehensively analyzed and discussed: failure rejection, unknown rejection, and fake rejection. By rejecting failures, the risk could be controlled especially for mission-critical applications. By rejecting unknowns, the awareness of the knowledge blind zone would be enhanced. By rejecting fakes, security and privacy could be protected. We provide a general taxonomy, organization, and discussion of the methods for solving these problems, which are studied separately in the literature. The connections between different approaches and future directions that are worth further investigation are also presented. With a discriminative and calibrated confidence, learning to reject will let the decision-making process be more practical, reliable, and secure.

Pretraining large models on generous multi-modal corpora has accelerated the development of visual-linguistic (VL) representation and achieved great success on various vision-and-language downstream tasks. Learning these models is usually executed by predicting the randomly masked words of captions or patches in images. Such approaches, nevertheless, seldom explore the supervision of causalities behind the caption descriptions or the procedure of generating events beyond still images. In this work, we endow the pretrained models with high-level cognition by delving into dynamic contexts to model the visual and linguistic causalities uniformly. Specifically, we format thedynamic contextsof an image as the sentences describing the eventsbefore,on, andafterimage. Unlike traditional caption-wise similarity, we propose a novel dynamic contexts-based similarity (DCS) metric, in which the correlation of potential causes and effects besides immediate visual content are considered to measure the relevance among images. DCS can be further simplified by parameterizing event continuity to relax the requirements on dense contextual event annotations. A new pre-task is designed to minimize the feature distances of dynamically contextual relevant images and incorporate the event causality and commonsense knowledge into the VL representation learning. Models based on our dynamic contexts significantly outperform typical VL models on multiple cross-modal downstream tasks, including the conventional visual commonsense reasoning (VCR), visual question answering (VQA), zero-shot image-text retrieval, and extended image / event ordering tasks.

Vision-and-Language Navigation (VLN) has been an emerging and fast-developing research topic, where an embodied agent is required to navigate in a real-world environment based on natural language instructions. In this article, we present a Direction-guided Navigator Agent (DNA) that novelly integrates direction clues derived from instructions into the essential encoder-decoder navigation framework. Particularly, DNA couples the standard instruction encoder with an additional direction branch which sequentially encodes the direction clues in the instructions to boost navigation. Furthermore, an Instruction Flipping mechanism is uniquely devised to enable fast data augmentation as well as a follow-up backtracing for navigating the agent in a backward direction. Such a way naturally amplifies the grounding of instruction in the local visual scenes along both forward and backward directions, and thus strengthens the alignment between instruction and action sequence. Extensive experiments conducted on Room to Room (R2R) dataset validate our proposal and demonstrate quantitatively compelling results.

Video captioning has been an emerging research topic in computer vision, which aims to generate a natural sentence to correctly reflect the visual content of a video. The well-established way of doing so is to rely on encoder-decoder paradigm by learning to encode the input video and decode the variable-length output sentence in a sequence-to-sequence manner. Nevertheless, these approaches often fail to produce complex and descriptive sentences as natural as those from human being, since the models are incapable of memorizing all visual contents and syntactic structures in the human-annotated video-sentence pairs. In this article, we uniquely introduce a Retrieval Augmentation Mechanism (RAM) that enables the explicit reference to existing video-sentence pairs within any encoder-decoder captioning model. Specifically, for each query video, a video-sentence retrieval model is first utilized to fetch semantically relevant sentences from the training sentence pool, coupled with the corresponding training videos. RAM then writes the relevant video-sentence pairs into memory and reads the memorized visual contents/syntactic structures in video-sentence pairs from memory to facilitate the word prediction at each timestep. Furthermore, we present Retrieval Augmented Convolutional Encoder-Decoder Network (R-ConvED), which novelly integrates RAM into convolutional encoder-decoder structure to boost video captioning. Extensive experiments on MSVD, MSR-VTT, Activity Net Captions, and VATEX datasets validate the superiority of our proposals and demonstrate quantitatively compelling results.

Visual and spatial relationship detection in images has been a fast-developing research topic in the multimedia field, which learns to recognize the semantic/spatial interactions between objects in an image, aiming to compose a structured semantic understanding of the scene. Most of the existing techniques directly encapsulate the holistic image feature plus the semantic and spatial features of the given two objects for predicting the relationship, but leave the inherent supervision derived from such structured and thorough image understanding under-exploited. Specifically, the inherent supervision among objects or relations within an image can span different granularities in this hierarchy including, from simple to comprehensive, (1) the object-based supervision that captures the interaction between the semantic and spatial features of each individual object, (2) the inter-object supervision that characterizes the dependency within the relationship triplet ( ), and (3) the inter-relation supervision that exploits contextual information among all relationship triplets in an image. These inherent multi-granular supervisions offer a fertile ground for building self-supervised proxy tasks. In this article, we compose a trilogy of exploring the multi-granular supervision in the sequence from object-based, inter-object, and inter-relation perspectives. We integrate the standard relationship detection objective with a series of proposed self-supervised proxy tasks, which is named as Multi-Granular Self-Supervised learning (MGS). Our MGS is appealing in view that it is pluggable to any neural relationship detection models by simply including the proxy tasks during training, without increasing the computational cost at inference. Through extensive experiments conducted on the SpatialSense and VRD datasets, we demonstrate the superiority of MGS for both spatial and visual relationship detection tasks.

A bottom-up and top-down attention mechanism has led to the revolutionizing of image captioning techniques, which enables object-level attention for multi-step reasoning over all the detected objects. However, when humans describe an image, they often apply their own subjective experience to focus on only a few salient objects that are worthy of mention, rather than all objects in this image. The focused objects are further allocated in linguistic order, yielding the “object sequence of interest” to compose an enriched description. In this work, we present the Bottom-up and Top-down Object inference Network (BTO-Net), which novelly exploits the object sequence of interest as top-down signals to guide image captioning. Technically, conditioned on the bottom-up signals (all detected objects), an LSTM-based object inference module is first learned to produce the object sequence of interest, which acts as the top-down prior to mimic the subjective experience of humans. Next, both of the bottom-up and top-down signals are dynamically integrated via an attention mechanism for sentence generation. Furthermore, to prevent the cacophony of intermixed cross-modal signals, a contrastive learning-based objective is involved to restrict the interaction between bottom-up and top-down signals, and thus leads to reliable and explainable cross-modal reasoning. Our BTO-Net obtains competitive performances on the COCO benchmark, in particular, 134.1% CIDEr on the COCO Karpathy test split. Source code is available at https://github.com/YehLi/BTO-Net .

With the recent development of deep convolutional neural networks and large-scale datasets, deep face recognition has made remarkable progress and been widely used in various applications. However, unlike the existing public face datasets, in many real-world scenarios of face recognition, the depth of the training dataset is shallow, which means that only two face images are available for each ID. With the non-uniform increase of samples, such issue is converted to a more general case, known as long-tail face learning, which suffers from data imbalance and intra-class diversity dearth simultaneously. These adverse conditions damage the training and result in the decline of model performance. Based on Semi-Siamese Training, we introduce an advanced solution, namedMulti-Agent Semi-Siamese Training(MASST), to address these problems. MASST includes a probe network and multiple gallery agents—the former aims to encode the probe features, and the latter constitutes a stack of networks that encode the prototypes (gallery features). For each training iteration, the gallery network, which is sequentially rotated from the stack, and the probe network form a pair of Semi-Siamese networks. We give the theoretical and empirical analysis that, given the long-tail (or shallow) data and training loss, MASST smooths the loss landscape and satisfies the Lipschitz continuity with the help of multiple agents and the updating gallery queue. The proposed method is out of extra-dependency, and thus can be easily integrated with the existing loss functions and network architectures. It is worth noting that although multiple gallery agents are employed for training, only the probe network is needed for inference, without increasing the inference cost. Extensive experiments and comparisons demonstrate the advantages of MASST for long-tail and shallow face learning.

The scene graph is a symbolic data structure that comprehensively describes the objects and visual relations in a visual scene, while ignoring the inherent perceptual saliency of each visual relation (i.e., relation saliency). However, humans often quickly allocate attention to important/salient visual relations in a scene. To align with such human perception of a scene, we explicitly model the perceptual saliency of visual relation in scene graph by upgrading each graph edge (i.e., visual relation) with an attribute of relation saliency. We present a new design, named as Saliency-guided Message Passing (SMP), that boosts the generation of such scene graph structure with the guidance from the visual relation saliency. Technically, an object interaction encoder is first utilized to strengthen object relation representations by jointly exploiting the appearance, semantic, and spatial relations in between. A branch is further leveraged to estimate the relation saliency of each visual relation by ordinal regression. Next, conditioned on the object and relation features (coupled with the estimated relation saliency), our SMP enhances scene graph generation by performing message passing over the objects and the most salient relations. Extensive experiments on VG-KR and VG150 datasets demonstrate the superiority of SMP for the scene graph generation. Moreover, we empirically validate the compelling generalizability of the learned scene graphs via SMP on downstream tasks like cross-model retrieval and image captioning.

Despite the large augmentation family, only a few cherry-picked robust augmentation policies are beneficial to self-supervised image representation learning. In this paper, we propose a directional self-supervised learning paradigm (DSSL), which is compatible with significantly more augmentations. Specifically, we adapt heavy augmentation policies after the views lightly augmented by standard augmentations, to generate harder view (HV). HV usually has a higher deviation from the original image than the lightly augmented standard view (SV). Unlike previous methods equally pairing all augmented views to symmetrically maximize their similarities, DSSL treats augmented views of the same instance as a partially ordered set (with directions as SV↔SV, SV↔HV), and then equips a directional objective function respecting to the derived relationships among views. DSSL can be easily implemented with a few lines of codes and is highly flexible to popular self-supervised learning frameworks, including SimCLR, Sim-Siam, BYOL. Extensive experimental results on CIFAR and ImageNet demonstrated that DSSL can stably improve various baselines with compatibility to a wider range of augmentations. Code is available at: https://github.com/Yif-Yang/DSSL.

Comprehending the rich semantics in an image and ordering them in linguistic order are essential to compose a visually-grounded and linguistically coherent description for image captioning. Modern techniques commonly capitalize on a pre-trained object detector/classifier to mine the semantics in an image, while leaving the inherent linguistic ordering of semantics under-exploited. In this paper, we propose a new recipe of Transformer-style structure, namely Comprehending and Ordering Semantics Networks (COS-Net), that novelly unifies an enriched semantic comprehending and a learnable semantic ordering processes into a single architecture. Technically, we initially utilize a cross-modal retrieval model to search the relevant sentences of each image, and all words in the searched sentences are taken as primary semantic cues. Next, a novel semantic comprehender is devised to filter out the irrelevant semantic words in primary semantic cues, and mean-while infer the missing relevant semantic words visually grounded in the image. After that, we feed all the screened and enriched semantic words into a semantic ranker, which learns to allocate all semantic words in linguistic order as humans. Such sequence of ordered semantic words are further integrated with visual tokens of images to trigger sentence generation. Empirical evidences show that COS-Net clearly surpasses the state-of-the-art approaches on COCO and achieves to-date the best CIDEr score of 141.1% on Karpathy test split. Source code is available at https://github.com/YehLi/xmodaler/tree/master/configs/image_caption/cosnet.

Motion, as the uniqueness of a video, has been critical to the development of video understanding models. Modern deep learning models leverage motion by either executing spatio-temporal 3D convolutions, factorizing 3D convolutions into spatial and temporal convolutions separately, or computing self-attention along temporal dimension. The implicit assumption behind such successes is that the feature maps across consecutive frames can be nicely aggregated. Nevertheless, the assumption may not always hold especially for the regions with large deformation. In this paper, we present a new recipe of inter-frame attention block, namely Stand-alone Inter-Frame Attention (SIFA), that novelly delves into the deformation across frames to estimate local self-attention on each spatial location. Technically, SIFA remoulds the deformable design via re-scaling the offset predictions by the difference between two frames. Taking each spatial location in the current frame as the query, the locally deformable neighbors in the next frame are regarded as the keys/values. Then, SIFA measures the similarity between query and keys as stand-alone attention to weighted average the values for temporal aggregation. We further plug SIFA block into ConvNets and Vision Transformer, respectively, to devise SIFA-Net and SIFA-Transformer. Extensive experiments conducted on four video datasets demonstrate the superiority of SIFA-Net and SIFA-Transformer as stronger backbones. More remarkably, SIFA-Transformer achieves an accuracy of 83.1% on Kinetics-400 dataset. Source code is available at https://github.com/FuchenUSTC/SIFA.

Convolutional Neural Networks (CNNs) have been re-garded as the go-to models for visual recognition. More re-cently, convolution-free networks, based on multi-head self-attention (MSA) or multi-layer perceptrons (MLPs), become more and more popular. Nevertheless, it is not trivial when utilizing these newly-minted networks for video recognition due to the large variations and complexities in video data. In this paper, we present MLP-3D networks, a novel MLP-like 3D architecture for video recognition. Specifically, the architecture consists of MLP-3D blocks, where each block contains one MLP applied across tokens (i.e., token-mixing MLP) and one MLP applied independently to each token (i.e., channel MLP). By deriving the novel grouped time mixing (GTM) operations, we equip the basic token-mixing MLP with the ability of temporal modeling. GTM divides the input tokens into several temporal groups and linearly maps the tokens in each group with the shared projection matrix. Furthermore, we devise several variants of GTM with different grouping strategies, and compose each vari-ant in different blocks of MLP-3D network by greedy ar-chitecture search. Without the dependence on convolutions or attention mechanisms, our MLP-3D networks achieves 68.5%/81.4% top-1 accuracy on Something-Something V2 and Kinetics-400 datasets, respectively. Despite with fewer computations, the results are comparable to state-of-the-art widely-used 3D CNNs and video transformers.

Given an image with multiple people, our goal is to directly regress the pose and shape of all the people as well as their relative depth. Inferring the depth of a person in an image, however, is fundamentally ambiguous without knowing their height. This is particularly problematic when the scene contains people of very different sizes, e.g. from infants to adults. To solve this, we need several things. First, we develop a novel method to infer the poses and depth of multiple people in a single image. While previous work that estimates multiple people does so by reasoning in the image plane, our method, called BEV, adds an additional imaginary Bird's-Eye-View representation to explicitly reason about depth. BEV reasons simultaneously about body centers in the image and in depth and, by combing these, estimates 3D body position. Unlike prior work, BEV is a single-shot method that is end-to-end differentiable. Second, height varies with age, making it impossible to resolve depth without also estimating the age of people in the image. To do so, we exploit a 3D body model space that lets BEV infer shapes from infants to adults. Third, to train BEV, we need a new dataset. Specifically, we create a “Relative Human” (RH) dataset that includes age labels and relative depth relationships between the people in the images. Extensive experiments on RH and AGORA demonstrate the effectiveness of the model and training scheme. BEV out-performs existing methods on depth reasoning, child shape estimation, and robustness to occlusion. The code11https://github.com/Arthur151/ROMP and dataset22https://github.com/Arthur151/Relative_Human are released for research purposes.

In this paper, we propose a simple yet effective video super-resolution method that aims at generating highfidelity high-resolution (HR) videos from low-resolution (LR) ones. Previous methods predominantly leverage temporal neighbor frames to assist the super-resolution of the current frame. Those methods achieve limited performance as they suffer from the challenges in spatial frame alignment and the lack of useful information from similar LR neighbor frames. In contrast, we devise a cross-frame non-local attention mechanism that allows video superresolution without frame alignment, leading to being more robust to large motions in the video. In addition, to acquire general video prior information beyond neighbor frames, and to compensate for the information loss caused by large motions, we design a novel memory-augmented attention module to memorize general video details during the superresolution training. We have thoroughly evaluated our work on various challenging datasets. Compared to other recent video super-resolution approaches, our method not only achieves significant performance gains on large motion videos but also shows better generalization. Our source code and the new Parkour benchmark dataset is available at https://github.com/jiy173/MANA.

Recent high-performing Human-Object Interaction (HOI) detection techniques have been highly influenced by Transformer-based object detector (i.e., DETR). Nevertheless, most of them directly map parametric interaction queries into a set of HOI predictions through vanilla Transformer in a one-stage manner. This leaves rich interor intra-interaction structure under-exploited. In this work, we design a novel Transformer-style HOI detector, i.e., Structure-aware Transformer over Interaction Proposals (STIP), for HOI detection. Such design decomposes the process of HOI set prediction into two subsequent phases, i.e., an interaction proposal generation is first performed, and then followed by transforming the non-parametric interaction proposals into HOI predictions via a structure-aware Transformer. The structure-aware Transformer upgrades vanilla Transformer by encoding additionally the holistically semantic structure among interaction proposals as well as the locally spatial structure of human/object within each interaction proposal, so as to strengthen HOI predictions. Extensive experiments conducted on V-COCO and HICO-DET benchmarks have demonstrated the effectiveness of STIP, and superior results are reported when comparing with the state-of-the-art HOI detectors. Source code is available at https://github.com/zyong812/STIP.

Existing studies for gait recognition are dominated by 2D representations like the silhouette or skeleton of the human body in constrained scenes. However, humans live and walk in the unconstrained 3D space, so projecting the 3D human body onto the 2D plane will discard a lot of crucial information like the viewpoint, shape, and dynamics for gait recognition. Therefore, this paper aims to explore dense 3D representations for gait recognition in the wild, which is a practical yet neglected problem. In particular, we propose a novel framework to explore the 3D Skinned Multi-Person Linear (SMPL) model of the human body for gait recognition, named SMPLGait. Our framework has two elaborately-designed branches of which one extracts appearance features from silhouettes, the other learns knowledge of 3D viewpoints and shapes from the 3D SMPL model. In addition, due to the lack of suitable datasets, we build the first large-scale 3D representation-based gait recognition dataset, named Gait3D. It contains 4,000 subjects and over 25,000 sequences extracted from 39 cameras in an unconstrained indoor scene. More importantly, it provides 3D SMPL models recovered from video frames which can provide dense 3D information of body shape, viewpoint, and dynamics. Based on Gait3D, we comprehensively compare our method with existing gait recognition approaches, which reflects the superior performance of our framework and the potential of 3D representations for gait recognition in the wild. The code and dataset are available at: https://gait3d.github.io.

In this study, we aim to predict the plausible future action steps given an observation of the past and study the task of instructional activity anticipation. Unlike previous anticipation tasks that aim at action label prediction, our work targets at generating natural language outputs that provide interpretable and accurate descriptions of future action steps. It is a challenging task due to the lack of semantic information extracted from the instructional videos. To overcome this challenge, we propose a novel knowledge distillation framework to exploit the related external textual knowledge to assist the visual anticipation task. However, previous knowledge distillation techniques generally transfer information within the same modality. To bridge the gap between the visual and text modalities during the distillation process, we devise a novel cross-modal contrastive distillation (CCD) scheme, which facilitates knowledge distillation between teacher and student in heterogeneous modalities with the proposed cross-modal distillation loss. We evaluate our method on the Tasty Videos dataset. CCD improves the anticipation performance of the visual-alone student model by a large margin of 40.2% relatively in BLEU4. Our approach also outperforms the state-of-the-art approaches by a large margin.

Action recognition from videos, i.e., classifying a video into one of the pre-defined action types, has been a popular topic in the communities of artificial intelligence, multimedia, and signal processing. However, existing methods usually consider an input video as a whole and learn models, e.g., Convolutional Neural Networks (CNNs), with coarse video-level class labels. These methods can only output an action class for the video, but cannot provide fine-grained and explainable cues to answer why the video shows a specific action. Therefore, researchers start to focus on a new task, Part-level Action Parsing (PAP), which aims to not only predict the video-level action but also recognize the frame-level fine-grained actions or interactions of body parts for each person in the video. To this end, we propose a coarse-to-fine framework for this challenging task. In particular, our framework first predicts the video-level class of the input video, then localizes the body parts and predicts the part-level action. Moreover, to balance the accuracy and computation in part-level action parsing, we propose to recognize the part-level actions by segment-level features. Furthermore, to overcome the ambiguity of body parts, we propose a pose-guided positional embedding method to accurately localize body parts. Through comprehensive experiments on a large-scale dataset, i.e., Kinetics-TPS, our framework achieves state-of-the-art performance and outperforms existing methods over 31.10% ROC score.

Recognizing human actions from point cloud videos has attracted tremendous attention from both academia and industry due to its wide applications like automatic driving, robotics, and so on. However, current methods for point cloud action recognition usually require a huge amount of data with manual annotations and a complex backbone network with high computation cost, which makes it impractical for real-world applications. Therefore, this paper considers the task of semi-supervised point cloud action recognition. We propose a Masked Pseudo-Labeling autoEncoder (MAPLE) framework to learn effective representations with much fewer annotations for point cloud action recognition. In particular, we design a novel and efficient Decoupled spatial-temporal TransFormer (DestFormer) as the backbone of MAPLE. In DestFormer, the spatial and temporal dimensions of the 4D point cloud videos are decoupled to achieve an efficient self-attention for learning both long-term and short-term features. Moreover, to learn discriminative features from fewer annotations, we design a masked pseudo-labeling autoencoder structure to guide the DestFormer to reconstruct features of masked frames from the available frames. More importantly, for unlabeled data, we exploit the pseudo-labels from the classification head as the supervision signal for the reconstruction of features from the masked frames. Finally, comprehensive experiments demonstrate that MAPLE achieves superior results on three public benchmarks and outperforms the state-of-the-art method by 8.08% accuracy on the MSR-Action3D dataset.

In this work, we present Auto-captions on GIF (ACTION), which is a new large-scale pre-training dataset for generic video understanding. All video-sentence pairs are created by automatically extracting and filtering video caption annotations from billions of web pages. Auto-captions on GIF dataset can be utilized to pre-train the generic feature representation or encoder-decoder structure for video captioning, and other downstream tasks (e.g., sentence localization in videos, video question answering, etc.) as well. We present a detailed analysis of Auto-captions on GIF dataset in comparison to existing video-sentence datasets. We also provide an evaluation of a Transformer-based encoder-decoder structure for vision-language pre-training, which is further adapted to video captioning downstream task and yields the compelling generalizability on MSR-VTT. The dataset is available at http://www.auto-video-captions.top/2022/dataset.

We develop WOC, a webcam-based 3D virtual online chatroom for multi-person interaction, which captures the 3D motion of users and drives their individual 3D virtual avatars in real-time. Compared to the existing wearable equipment-based solution, WOC offers convenient and low-cost 3D motion capture with a single camera. To promote the immersive chat experience, WOC provides high-fidelity virtual avatar manipulation, which also supports the user-defined characters. With the distributed data flow service, the system delivers highly synchronized motion and voice for all users. Deployed on the website and no installation required, users can freely experience the virtual online chat at https://yanch.cloud/.

Recently, various methods have been introduced to address the OOD detection problem with training outlier exposure. These methods usually count on discriminative softmax metric or energy method to screen OOD samples. In this paper, we probe an alternative hypothesis on OOD detection by constructing a novel latent variable model based on independent component analysis (ICA) techniques. This novel method named Conditional-i builds upon the probabilistic formulation, and applies the Hilbert-Schmidt Independence Criteria that offers a convenient solution for optimizing variable dependencies. Conditional-i exclusively encodes the useful class condition into the probabilistic model, which provides the desired convenience in delivering theoretical support for the OOD detection task. To facilitate the implementation of the Conditional-i model, we construct unique memory bank architectures that allow for convenient end-to-end training within a tractable budget. Empirical results demonstrate an evident performance boost on benchmarks against SOTA methods. We also provide valuable theoretical justifications that our training strategy is guaranteed to bound the error in the context of OOD detection. Code is available at: https://github.com/OODHSIC/conditional-i.

The adaptation of a Generative Adversarial Network (GAN) aims to transfer a pre-trained GAN to a target domain with limited training data. In this paper, we focus on the one-shot case, which is more challenging and rarely explored in previous works. We consider that the adaptation from a source domain to a target domain can be decoupled into two parts: the transfer of global style like texture and color, and the emergence of new entities that do not belong to the source domain. While previous works mainly focus on style transfer, we propose a novel and concise framework to address the \textit{generalized one-shot adaptation} task for both style and entity transfer, in which a reference image and its binary entity mask are provided. Our core idea is to constrain the gap between the internal distributions of the reference and syntheses by sliced Wasserstein distance. To better achieve it, style fixation is used at first to roughly obtain the exemplary style, and an auxiliary network is introduced to the generator to disentangle entity and style transfer. Besides, to realize cross-domain correspondence, we propose the variational Laplacian regularization to constrain the smoothness of the adapted generator. Both quantitative and qualitative experiments demonstrate the effectiveness of our method in various scenarios. Code is available at \url{https://github.com/zhangzc21/Generalized-One-shot-GAN-adaptation}.

Although deep face recognition benefits significantly from large-scale training data, a current bottleneck is the labelling cost. A feasible solution to this problem is semi-supervised learning, exploiting a small portion of labelled data and large amounts of unlabelled data. The major challenge, however, is the accumulated label errors through auto-labelling, compromising the training. In this paper, we present an effective solution to semi-supervised face recognition that is robust to the label noise aroused by the auto-labelling. Specifically, we introduce a multi-agent method, named GroupNet (GN), to endow our solution with the ability to identify the wrongly-labelled samples and preserve the clean samples. We show that GN alone achieves the leading accuracy in traditional supervised face recognition even when the noisy labels take over 50% of the training data. Further, we develop a semi-supervised face recognition solution, named Noise Robust Learning-Labelling (NRoLL), which is based on the robust training ability empowered by GN. It starts with a small amount of labelled data and consequently conducts high-confidence labelling on a large amount of unlabelled data to boost further training. The more data is labelled by NRoLL, the higher confidence is with the label in the dataset. To evaluate the competitiveness of our method, we run NRoLL with a rough condition that only one-fifth of the labelled MSCeleb is available and the rest is used as unlabelled data. On a wide range of benchmarks, our method compares favorably against the state-of-the-art methods.

Computer Vision (CV) and Natural Language Processing (NLP) are two most fundamental disciplines under a broad area of artificial intelligence (AI). CV is regarded as a field of research that explores the techniques to teach computers to see and understand digital content such as images and videos. NLP is a branch of linguistics that enables computers to process, interpret, and even generate human language. With the rise and development of deep learning over the past decade, there has been a steady momentum of innovation and breakthroughs that convincingly push the limits and improve the state-of-the-art of both vision and language modeling. An interesting observation is that the research in the two areas starts to interact, with a significant growth in both the volume of publications and extensive applications. Meanwhile, many previous experiences have shown that this can naturally build up the circle of human intelligence.

Face anti-spoofing (FAS) plays a vital role in preventing face recognition systems from presentation attacks. Existing face anti-spoofing datasets lack diversity due to the insufficient identity and insignificant variance, which limits the generalization ability of FAS model. In this paper, we propose Dual Spoof Disentanglement Generation (DSDG) framework to tackle this challenge by “anti-spoofing via generation”. Depending on the interpretable factorized latent disentanglement in Variational Autoencoder (VAE), DSDG learns a joint distribution of the identity representation and the spoofing pattern representation in the latent space. Then, large-scale paired live and spoofing images can be generated from random noise to boost the diversity of the training set. However, some generated face images are partially distorted due to the inherent defect of VAE. Such noisy samples are hard to predict precise depth values, thus may obstruct the widely-used depth supervised optimization. To tackle this issue, we further introduce a lightweight Depth Uncertainty Module (DUM), which alleviates the adverse effects of noisy samples by depth uncertainty learning. DUM is developed without extra-dependency, thus can be flexibly integrated with any depth supervised network for face anti-spoofing. We evaluate the effectiveness of the proposed method on five popular benchmarks and achieve state-of-the-art results under both intra- and inter- test settings. The codes are available athttps://github.com/JDAI-CV/FaceX-Zoo/tree/main/addition_module/DSDG.

Recent progress on 2D object detection has featured Cascade RCNN, which capitalizes on a sequence of cascade detectors to progressively improve proposal quality, towards high-quality object detection. However, there has not been evidence in support of building such cascade structures for 3D object detection, a challenging detection scenario with highly sparse LiDAR point clouds. In this work, we present a simple yet effective cascade architecture, named 3D Cascade RCNN, that allocates multiple detectors based on the voxelized point clouds in a cascade paradigm, pursuing higher quality 3D object detector progressively. Furthermore, we quantitatively define the sparsity level of the points within 3D bounding box of each object as the point completeness score, which is exploited as the task weight for each proposal to guide the learning of each stage detector. The spirit behind is to assign higher weights for high-quality proposals with relatively complete point distribution, while down-weight the proposals with extremely sparse points that often incur noise during training. This design of completeness-aware re-weighting elegantly upgrades the cascade paradigm to be better applicable for the sparse input data, without increasing any FLOP budgets. Through extensive experiments on both the KITTI dataset and Waymo Open Dataset, we validate the superiority of our proposed 3D Cascade RCNN, when comparing to state-of-the-art 3D object detection techniques. The source code is publicly available at https://github.com/caiqi/Cascasde-3D.

Image captioning has been an emerging and fast-developing research topic. Nevertheless, most existing works heavily rely on large amounts of image-sentence pairs and therefore hinder the practical applications of captioning in the wild. In this paper, we present a novel Semantic-Constrained Self-learning (SCS) framework that explores an iterative self-learning strategy to learn an image captioner with only unpaired image and text data. Technically, SCS consists of two stages, i.e., pseudo pair generation and captioner re-training, iteratively producing "pseudo" image-sentence pairs via a pre-trained captioner and re-training the captioner with the pseudo pairs, respectively. Particularly, both stages are guided by the recognized objects in the image, that act as semantic constraint to strengthen the semantic alignment between the input image and the output sentence. We leverage a semantic-constrained beam search for pseudo pair generation to regularize the decoding process with the recognized objects via forcing the inclusion/exclusion of the recognized/irrelevant objects in output sentence. For captioner re-training, a self-supervised triplet loss is utilized to preserve the relative semantic similarity ordering among generated sentences with regard to the input image triplets. Moreover, an object inclusion reward and an adversarial reward are adopted to encourage the inclusion of the predicted objects in the output sentence and pursue the generation of more realistic sentences during self-critical training, respectively. Experiments conducted on both dependent and independent unpaired data validate the superiority of SCS. More remarkably, we obtain the best published CIDEr score to-date of 74.7\% on COCO Karpathy test split for unpaired image captioning.

The performance of human pose estimation depends on the spatial accuracy of keypoint localization. Most existing methods pursue the spatial accuracy through learning the high-resolution (HR) representation from input images. By the experimental analysis, we find that the HR representation leads to a sharp increase of computational cost, while the accuracy improvement remains marginal compared with the low-resolution (LR) representation. In this article, we propose a design paradigm for cost-effective network with LR representation for efficient pose estimation, named FasterPose. Whereas the LR design largely shrinks the model complexity, how to effectively train the network with respect to the spatial accuracy is a concomitant challenge. We study the training behavior of FasterPose and formulate a novel regressive cross-entropy (RCE) loss function for accelerating the convergence and promoting the accuracy. The RCE loss generalizes the ordinary cross-entropy loss from the binary supervision to a continuous range, thus the training of pose estimation network is able to benefit from the sigmoid function. By doing so, the output heatmap can be inferred from the LR features without loss of spatial accuracy, while the computational cost and model size has been significantly reduced. Compared with the previously dominant network of pose estimation, our method reduces 58% of the FLOPs and simultaneously gains 1.3% improvement of accuracy. Extensive experiments show that FasterPose yields promising results on the common benchmarks, i.e., COCO and MPII, consistently validating the effectiveness and efficiency for practical utilization, especially the low-latency and low-energy-budget applications in the non-GPU scenarios.

Vision-language pre-training has been an emerging and fast-developing research topic, which transfers multi-modal knowledge from rich-resource pre-training task to limited-resource downstream tasks. Unlike existing works that predominantly learn a single generic encoder, we present a pre-trainable Universal Encoder-DEcoder Network (Uni-EDEN) to facilitate both vision-language perception (e.g., visual question answering) and generation (e.g., image captioning). Uni-EDEN is a two-stream Transformer-based structure, consisting of three modules: object and sentence encoders that separately learns the representations of each modality and sentence decoder that enables both multi-modal reasoning and sentence generation via inter-modal interaction. Considering that the linguistic representations of each image can span different granularities in this hierarchy including, from simple to comprehensive, individual label, a phrase, and a natural sentence, we pre-train Uni-EDEN through multi-granular vision-language proxy tasks: Masked Object Classification, Masked Region Phrase Generation, Image-Sentence Matching, and Masked Sentence Generation. In this way, Uni-EDEN is endowed with the power of both multi-modal representation extraction and language modeling. Extensive experiments demonstrate the compelling generalizability of Uni-EDEN by fine-tuning it to four vision-language perception and generation downstream tasks.

The significant progress on Generative Adversarial Networks (GANs) has facilitated realistic single-object image generation based on language input. However, complex-scene generation (with various interactions among multiple objects) still suffers from messy layouts and object distortions, due to diverse configurations in layouts and appearances. Prior methods are mostly object-driven and ignore their inter-relations that play a significant role in complex-scene images. This work explores relationship-aware complex-scene image generation, where multiple objects are inter-related as a scene graph. With the help of relationships, we propose three major updates in the generation framework. First, reasonable spatial layouts are inferred by jointly considering the semantics and relationships among objects. Compared to standard location regression, we show relative scales and distances serve a more reliable target. Second, since the relations between objects have significantly influenced an object's appearance, we design a relation-guided generator to generate objects reflecting their relationships. Third, a novel scene graph discriminator is proposed to guarantee the consistency between the generated image and the input scene graph. Our method tends to synthesize plausible layouts and objects, respecting the interplay of multiple objects in an image. Experimental results on Visual Genome and HICO-DET datasets show that our proposed method significantly outperforms prior arts in terms of IS and FID metrics. Based on our user study and visual inspection, our method is more effective in generating logical layout and appearance for complex-scenes.

Despite having impressive vision-language (VL) pretraining with BERT-based encoder for VL understanding, the pretraining of a universal encoder-decoder for both VL understanding and generation remains challenging. The difficulty originates from the inherently different peculiarities of the two disciplines, e.g., VL understanding tasks capitalize on the unrestricted message passing across modalities, while generation tasks only employ visual-to-textual message passing. In this paper, we start with a two-stream decoupled design of encoder-decoder structure, in which two decoupled cross-modal encoder and decoder are involved to separately perform each type of proxy tasks, for simultaneous VL understanding and generation pretraining. Moreover, for VL pretraining, the dominant way is to replace some input visual/word tokens with mask tokens and enforce the multi-modal encoder/decoder to reconstruct the original tokens, but no mask token is involved when fine-tuning on downstream tasks. As an alternative, we propose a primary scheduled sampling strategy that elegantly mitigates such discrepancy via pretraining encoder-decoder in a two-pass manner. Extensive experiments demonstrate the compelling generalizability of our pretrained encoder-decoder by fine-tuning on four VL understanding and generation downstream tasks. Source code is available at https://github.com/YehLi/TDEN.

A steady momentum of innovations and breakthroughs has convincingly pushed the limits of unsupervised image representation learning. Compared to static 2D images, video has one more dimension (time). The inherent supervision existing in such sequential structure offers a fertile ground for building unsupervised learning models. In this paper, we compose a trilogy of exploring the basic and generic supervision in the sequence from spatial, spatiotemporal and sequential perspectives. We materialize the supervisory signals through determining whether a pair of samples is from one frame or from one video, and whether a triplet of samples is in the correct temporal order. We uniquely regard the signals as the foundation in contrastive learning and derive a particular form named Sequence Contrastive Learning (SeCo). SeCo shows superior results under the linear protocol on action recognition (Kinetics), untrimmed activity recognition (ActivityNet) and object tracking (OTB-100). More remarkably, SeCo demonstrates considerable improvements over recent unsupervised pre-training techniques, and leads the accuracy by 2.96% and 6.47% against fully-supervised ImageNet pre-training in action recognition task on UCF101 and HMDB51, respectively. Source code is available at https://github.com/YihengZhang-CV/SeCo-Sequence-Contrastive-Learning.

Existing methods for large-scale point cloud semantic segmentation require expensive, tedious and error-prone manual point-wise annotation. Intuitively, weakly supervised training is a direct solution to reduce the labeling costs. However, for weakly supervised large-scale point cloud semantic segmentation, too few annotations will inevitably lead to ineffective learning of network. We propose an effective weakly supervised method containing two components to solve the above problem. Firstly, we construct a pretext task, \textit{i.e.,} point cloud colorization, with a self-supervised training manner to transfer the learned prior knowledge from a large amount of unlabeled point cloud to a weakly supervised network. In this way, the representation capability of the weakly supervised network can be improved by knowledge from a heterogeneous task. Besides, to generative pseudo label for unlabeled data, a sparse label propagation mechanism is proposed with the help of generated class prototypes, which is used to measure the classification confidence of unlabeled point. Our method is evaluated on large-scale point cloud datasets with different scenarios including indoor and outdoor. The experimental results show the large gain against existing weakly supervised methods and comparable results to fully supervised methods.

Human actions are typically of combinatorial structures or patterns, i.e., subjects, objects, plus spatio-temporal interactions in between. Discovering such structures is therefore a rewarding way to reason about the dynamics of interactions and recognize the actions. In this paper, we introduce a new design of sub-graphs to represent and encode the discriminative patterns of each action in the videos. Specifically, we present MUlti-scale Sub-graph LEarning (MUSLE) framework that novelly builds space-time graphs and clusters the graphs into compact sub-graphs on each scale with respect to the number of nodes. Technically, MUSLE produces 3D bounding boxes, i.e., tubelets, in each video clip, as graph nodes and takes dense connectivity as graph edges between tubelets. For each action category, we execute online clustering to decompose the graph into sub-graphs on each scale through learning Gaussian Mixture Layer and select the discriminative sub-graphs as action prototypes for recognition. Extensive experiments are conducted on both Something-Something V1 & V2 and Kinetics-400 datasets, and superior results are reported when comparing to state-of-the-art methods. More remarkably, our MUSLE achieves to-date the best reported accuracy of 65.0% on Something-Something V2 validation set.

Weakly supervised temporal action localization aims to detect and localize actions in untrimmed videos with only video-level labels during training. However, without frame-level annotations, it is challenging to achieve localization completeness and relieve background interference. In this paper, we present an Action Unit Memory Network (AUMN) for weakly supervised temporal action localization, which can mitigate the above two challenges by learning an action unit memory bank. In the proposed AUMN, two attention modules are designed to update the memory bank adaptively and learn action units specific classifiers. Furthermore, three effective mechanisms (diversity, homogeneity and sparsity) are designed to guide the updating of the memory network. To the best of our knowledge, this is the first work to explicitly model the action units with a memory network. Extensive experimental results on two standard benchmarks (THUMOS14 and ActivityNet) demonstrate that our AUMN performs favorably against state-of-the-art methods. Specifically, the average mAP of IoU thresholds from 0.1 to 0.5 on the THUMOS14 dataset is significantly improved from 47.0% to 52.1%.

Video content is multifaceted, consisting of objects, scenes, interactions or actions. The existing datasets mostly label only one of the facets for model training, resulting in the video representation that biases to only one facet depending on the training dataset. There is no study yet on how to learn a video representation from multifaceted labels, and whether multifaceted information is helpful for video representation learning. In this paper, we propose a new learning framework, MUlti-Faceted Integration (MUFI), to aggregate facets from different datasets for learning a representation that could reflect the full spectrum of video content. Technically, MUFI formulates the problem as visual-semantic embedding learning, which explicitly maps video representation into a rich semantic embedding space, and jointly optimizes video representation from two perspectives. One is to capitalize on the intra-facet supervision between each video and its own label descriptions, and the second predicts the "semantic representation" of each video from the facets of other datasets as the inter-facet supervision. Extensive experiments demonstrate that learning 3D CNN via our MUFI framework on a union of four large-scale video datasets plus two image datasets leads to superior capability of video representation. The prelearnt 3D CNN with MUFI also shows clear improvements over other approaches on several downstream video applications. More remarkably, MUFI achieves 98.1%/80.9% on UCF101/HMDB51 for action recognition and 101.5% in terms of CIDEr-D score on MSVD for video captioning.

Due to the subjective annotation and the inherent interclass similarity of facial expressions, one of key challenges in Facial Expression Recognition (FER) is the annotation ambiguity. In this paper, we proposes a solution, named DMUE, to address the problem of annotation ambiguity from two perspectives: the latent Distribution Mining and the pairwise Uncertainty Estimation. For the former, an auxiliary multi-branch learning framework is introduced to better mine and describe the latent distribution in the label space. For the latter, the pairwise relationship of semantic feature between instances are fully exploited to estimate the ambiguity extent in the instance space. The proposed method is independent to the backbone architectures, and brings no extra burden for inference. The experiments are conducted on the popular real-world benchmarks and the synthetic noisy datasets. Either way, the proposed DMUE stably achieves leading performance.

Unsupervised Domain Adaptive (UDA) person re-identification (ReID) aims at adapting the model trained on a labeled source-domain dataset to a target-domain dataset without any further annotations. Most successful UDA-ReID approaches combine clustering-based pseudo-label prediction with representation learning and perform the two steps in an alternating fashion. However, offline interaction between these two steps may allow noisy pseudo labels to substantially hinder the capability of the model. In this paper, we propose a Group-aware Label Transfer (GLT) algorithm, which enables the online interaction and mutual promotion of pseudo-label prediction and representation learning. Specifically, a label transfer algorithm simultaneously uses pseudo labels to train the data while refining the pseudo labels as an online clustering algorithm. It treats the online label refinery problem as an optimal transport problem, which explores the minimum cost for assigning M samples to N pseudo labels. More importantly, we introduce a group-aware strategy to assign implicit attribute group IDs to samples. The combination of the online label refining algorithm and the group-aware strategy can better correct the noisy pseudo label in an online fashion and narrow down the search space of the target identity. The effectiveness of the proposed GLT is demonstrated by the experimental results (Rank-1 accuracy) for Market1501→DukeMTMC (82.0%) and DukeMTMC→Market1501 (92.2%), remarkably closing the gap between unsupervised and supervised performance on person re-identification.1

Mainstream state-of-the-art domain generalization algorithms tend to prioritize the assumption on semantic in-variance across domains. Meanwhile, the inherent intra-domain style invariance is usually underappreciated and put on the shelf. In this paper, we reveal that leveraging intra-domain style invariance is also of pivotal importance in improving the efficiency of domain generalization. We verify that it is critical for the network to be informative on what domain features are invariant and shared among in-stances, so that the network sharpens its understanding and improves its semantic discriminative ability. Correspondingly, we also propose a novel “jury” mechanism, which is particularly effective in learning useful semantic feature commonalities among domains. Our complete model called STEAM can be interpreted as a novel probabilistic graphical model, for which the implementation requires convenient constructions of two kinds of memory banks: semantic feature bank and style feature bank. Empirical results show that our proposed framework surpasses the state-of-the-art methods by clear margins.

Despite the great progress of person re-identification (ReID) with the adoption of Convolutional Neural Networks, current ReID models are opaque and only outputs a scalar distance between two persons. There are few methods providing users semantically understandable explanations for why two persons are the same one or not. In this paper, we propose a post-hoc method, named Attribute-guided Metric Distillation (AMD), to explain existing ReID models. This is the first method to explore attributes to answer: 1) what and where the attributes make two persons different, and 2) how much each attribute contributes to the difference. In AMD, we design a pluggable interpreter network for target models to generate quantitative contributions of attributes and visualize accurate attention maps of the most discriminative attributes. To achieve this goal, we propose a metric distillation loss by which the interpreter learns to decompose the distance of two persons into components of attributes with knowledge distilled from the target model. Moreover, we propose an attribute prior loss to make the interpreter generate attribute-guided attention maps and to eliminate biases caused by the imbalanced distribution of attributes. This loss can guide the interpreter to focus on the exclusive and discriminative attributes rather than the large-area but common attributes of two persons. Comprehensive experiments show that the interpreter can generate effective and intuitive explanations for varied models and generalize well under cross-domain settings. As a by-product, the accuracy of target models can be further improved with our interpreter.1

Visible-Infrared person re-identification (VI-ReID) aims to match cross-modality pedestrian images, breaking through the limitation of single-modality person ReID in dark environment. In order to mitigate the impact of large modality discrepancy, existing works manually design various two-stream architectures to separately learn modality-specific and modality-sharable representations. Such a manual design routine, however, highly depends on massive experiments and empirical practice, which is time consuming and labor intensive. In this paper, we systematically study the manually designed architectures, and identify that appropriately separating Batch Normalization (BN) layers is the key to bring a great boost towards cross-modality matching. Based on this observation, the essential objective is to find the optimal separation scheme for each BN layer. To this end, we propose a novel method, named Cross-Modality Neural Architecture Search (CM-NAS). It consists of a BN-oriented search space in which the standard optimization can be fulfilled subject to the cross-modality task. Equipped with the searched architecture, our method outperforms state-of-the-art counter-parts in both two benchmarks, improving the Rank-1/mAP by 6.70%/6.13% on SYSU-MM01 and by 12.17%/11.23% on RegDB. Code is released at https://github.com/JDAI-CV/CM-NAS.

Motion, as the most distinct phenomenon in a video to involve the changes over time, has been unique and critical to the development of video representation learning. In this paper, we ask the question: how important is the motion particularly for self-supervised video representation learning. To this end, we compose a duet of exploiting the motion for data augmentation and feature learning in the regime of contrastive learning. Specifically, we present a Motion-focused Contrastive Learning (MCL) method that regards such duet as the foundation. On one hand, MCL capitalizes on optical flow of each frame in a video to temporally and spatially sample the tubelets (i.e., sequences of associated frame patches across time) as data augmentations. On the other hand, MCL further aligns gradient maps of the convolutional layers to optical flow maps from spatial, temporal and spatio-temporal perspectives, in order to ground motion information in feature learning. Extensive experiments conducted on R(2+1)D backbone demonstrate the effectiveness of our MCL. On UCF101, the linear classifier trained on the representations learnt by MCL achieves 81.91% top-1 accuracy, outperforming ImageNet supervised pre-training by 6.78%. On Kinetics-400, MCL achieves 66.62% top-1 accuracy under the linear protocol.

Video is complex due to large variations in motion and rich content in fine-grained visual details. Abstracting useful information from such information-intensive media requires exhaustive computing resources. This paper studies a two-step alternative that first condenses the video sequence to an informative "frame" and then exploits off-the-shelf image recognition system on the synthetic frame. A valid question is how to define "useful information" and then distill it from a video sequence down to one synthetic frame. This paper presents a novel Informative Frame Synthesis (IFS) architecture that incorporates three objective tasks, i.e., appearance reconstruction, video categorization, motion estimation, and two regularizers, i.e., adversarial learning, color consistency. Each task equips the synthetic frame with one ability, while each regularizer enhances its visual quality. With these, by jointly learning the frame synthesis in an end-to-end manner, the generated frame is expected to encapsulate the required spatio-temporal information useful for video analysis. Extensive experiments are conducted on the large-scale Kinetics dataset. When comparing to baseline methods that map video sequence to a single image, IFS shows superior performance. More remarkably, IFS consistently demonstrates evident improvements on image-based 2D networks and clip-based 3D networks, and achieves comparable performance with the state-of-the-art methods with less computational cost.

This paper focuses on the regression of multiple 3D people from a single RGB image. Existing approaches predominantly follow a multi-stage pipeline that first detects people in bounding boxes and then independently regresses their 3D body meshes. In contrast, we propose to Regress all meshes in a One-stage fashion for Multiple 3D People (termed ROMP). The approach is conceptually simple, bounding box-free, and able to learn a per-pixel representation in an end-to-end manner. Our method simultaneously predicts a Body Center heatmap and a Mesh Parameter map, which can jointly describe the 3D body mesh on the pixel level. Through a body-center-guided sampling process, the body mesh parameters of all people in the image are easily extracted from the Mesh Parameter map. Equipped with such a fine-grained representation, our one-stage framework is free of the complex multi-stage process and more robust to occlusion. Compared with state-of-the-art methods, ROMP achieves superior performance on the challenging multi-person benchmarks, including 3DPW and CMU Panoptic. Experiments on crowded/occluded datasets demonstrate the robustness under various types of occlusion. The code, released at https://github.com/Arthur151/ROMP, is the first real-time implementation of monocular multi-person 3D mesh regression.

It is not trivial to optimally learn a 3D Convolutional Neural Networks (3D ConvNets) due to high complexity and various options of the training scheme. The most common hand-tuning process starts from learning 3D ConvNets using short video clips and then is followed by learning long-term temporal dependency using lengthy clips, while gradually decaying the learning rate from high to low as training progresses. The fact that such process comes along with several heuristic settings motivates the study to seek an optimal "path" to automate the entire training. In this paper, we decompose the path into a series of training "states" and specify the hyper-parameters, e.g., learning rate and the length of input clips, in each state. The estimation of the knee point on the performance-epoch curve triggers the transition from one state to another. We perform dynamic programming over all the candidate states to plan the optimal permutation of states, i.e., optimization path. Furthermore, we devise a new 3D ConvNets with a unique design of dual-head classifier to improve spatial and temporal discrimination. Extensive experiments on seven public video recognition benchmarks demonstrate the advantages of our proposal. With the optimization planning, our 3D ConvNets achieves superior results when comparing to the state-of-the-art recognition methods. More remarkably, we obtain the top-1 accuracy of 80.5% and 82.7% on Kinetics-400 and Kinetics-600 datasets, respectively.

Gait, i.e., the movement pattern of human limbs during locomotion, is a promising biometrie for identification of persons. Despite significant improvement in gait recognition with deep learning, existing studies still neglect a more practical but challenging scenario - unsupervised cross-domain gait recognition which aims to learn a model on a labeled dataset then adapt it to an unlabeled dataset. Due to the domain shift and class gap, directly applying a model trained on one source dataset to other target datasets usually obtains very poor results. Therefore, this paper proposes a Transferable Neighborhood Discovery (TraND) framework to bridge the domain gap for unsupervised cross-domain gait recognition. To learn effective prior knowledge for gait representation, we first adopt a backbone network pre- trained on the labeled source data in a supervised manner. Then we design an end-to-end trainable approach to automatically discover the confident neighborhoods of unlabeled samples in the latent space. During training, the class consistency indicator is adopted to select confident neighborhoods of samples based on their entropy measurements. Moreover, we explore a high- entropy-first neighbor selection strategy, which can effectively transfer prior knowledge to the target domain. Our method achieves the state-of-the-art results on two public datasets, i.e., CASIA-B and OU-LP.

Self-supervised learning (SSL) has recently become the favorite among feature learning methodologies. It is therefore appealing for domain adaptation approaches to consider incorporating SSL. The intuition is to enforce instance-level feature consistency such that the predictor becomes somehow invariant across domains. However, most existing SSL methods in the regime of domain adaptation usually are treated as standalone auxiliary components, leaving the signatures of domain adaptation unattended. Actually, the optimal region where the domain gap vanishes and the instance level constraint that SSL peruses may not coincide at all. From this point, we present a particular paradigm of self-supervised learning tailored for domain adaptation, i.e., Transferrable Contrastive Learning (TCL), which links the SSL and the desired cross-domain transferability congruently. We find contrastive learning intrinsically a suitable candidate for domain adaptation, as its instance invariance assumption can be conveniently promoted to cross-domain class-level invariance favored by domain adaptation tasks. Based on particular memory bank constructions and pseudo label strategies, TCL then penalizes cross-domain intra-class domain discrepancy between source and target through a clean and novel contrastive loss. The free lunch is, thanks to the incorporation of contrastive learning, TCL relies on a moving-averaged key encoder that naturally achieves a temporally ensembled version of pseudo labels for target data, which avoids pseudo label error propagation at no extra cost. TCL therefore efficiently reduces cross-domain gaps. Through extensive experiments on benchmarks (Office-Home, VisDA-2017, Digits-five, PACS and DomainNet) for both single-source and multi-source domain adaptation tasks, TCL has demonstrated state-of-the-art performances.

With the rise and development of deep learning over the past decade, there has been a steady momentum of innovation and breakthroughs that convincingly push the state-of-the-art of cross-modal analytics between vision and language in multimedia field. Nevertheless, there has not been an open-source codebase in support of training and deploying numerous neural network models for cross-modal analytics in a unified and modular fashion. In this work, we propose X-modaler --- a versatile and high-performance codebase that encapsulates the state-of-the-art cross-modal analytics into several general-purpose stages (e.g., pre-processing, encoder, cross-modal interaction, decoder, and decode strategy). Each stage is empowered with the functionality that covers a series of modules widely adopted in state-of-the-arts and allows seamless switching in between. This way naturally enables a flexible implementation of state-of-the-art algorithms for image captioning, video captioning, and vision-language pre-training, aiming to facilitate the rapid development of research community. Meanwhile, since the effective modular designs in several stages (e.g., cross-modal interaction) are shared across different vision-language tasks, X-modaler can be simply extended to power startup prototypes for other tasks in cross-modal analytics, including visual question answering, visual commonsense reasoning, and cross-modal retrieval. X-modaler is an Apache-licensed codebase, and its source codes, sample projects and pre-trained models are available on-line: https://github.com/YehLi/xmodaler.

Face swapping aims to synthesize a face image, in which the facial identity is well transplanted from the source image and the context (e.g., hairstyle, head posture, facial expression, lighting, and background) keeps consistent with the reference image. The prior work mainly accomplishes the task in two stages, i.e., generating the inner face with the source identity, and then stitching the generation with the complementary part of the reference image by image blending techniques. The blending mask, which is usually obtained by the additional face segmentation model, is a common practice towards photo-realistic face swapping. However, artifacts usually appear at the blending boundary, especially in areas occluded by the hair, eyeglasses, accessories, etc. To address this problem, rather than struggling with the blending mask in the two-stage routine, we develop a novel one-stage context and identity hallucination network, which learns a series of hallucination maps to softly divide the context areas and identity areas. For context areas, the features are fully utilized by a multi-level context encoder. For identity areas, we design a novel two-cascading AdaIN to transfer the identity while retaining the context. Besides, with the help of hallucination maps, we introduce an effectively improved reconstruction loss to utilize unlimited unpaired face images for training. Our network performs well on both context areas and identity areas without any dependency on post-processing. Extensive qualitative and quantitative experiments demonstrate the superiority of our network.

BERT-type structure has led to the revolution of vision-language pre-training and the achievement of state-of-the-art results on numerous vision-language downstream tasks. Existing solutions dominantly capitalize on the multi-modal inputs with mask tokens to trigger mask-based proxy pre-training tasks (e.g., masked language modeling and masked object/frame prediction). In this work, we argue that such masked inputs would inevitably introduce noise for cross-modal matching proxy task, and thus leave the inherent vision-language association under-explored. As an alternative, we derive a particular form of cross-modal proxy objective for video-language pre-training, i.e., Contrastive Cross-modal matching and denoising (CoCo). By viewing the masked frame/word sequences as the noisy augmentation of primary unmasked ones, CoCo strengthens video-language association by simultaneously pursuing inter-modal matching and intra-modal denoising between masked and unmasked inputs in a contrastive manner. Our CoCo proxy objective can be further integrated into any BERT-type encoder-decoder structure for video-language pre-training, named as Contrastive Cross-modal BERT (CoCo-BERT). We pre-train CoCo-BERT on TV dataset and a newly collected large-scale GIF video dataset (ACTION). Through extensive experiments over a wide range of downstream tasks (e.g., cross-modal retrieval, video question answering, and video captioning), we demonstrate the superiority of CoCo-BERT as a pre-trained structure.

The world has become multimodal. In addition to text, we have been sharing a huge amount of multimedia information in the form of images and videos on the Internet. The wide spread use of smart mobile devices has also changed the way we interact with the Internet. It is now natural for us to capture images and videos freely and use as part of a query, in addition to the traditional text and voices. These, along with the rapid advancements in multimedia, natural language processing, information retrieval, and conversation technologies, mean that it is time for us to explore multimodal conversation and its roles in search and recommendation. Multimodal conversation has the potential to help us to uncover and digest the huge amount of multimedia information and knowledge hidden within many systems. It also enables a natural 2-way interactions between humans and machines, with mutual benefits in enriching their respective knowledge. Finally, it opens up the possibilities of disrupting many existing applications and launching new innovative applications. This panel is timely and aims to explore this emerging trend, and discuss its potential benefits and pitfalls to society. The panel will also explore the limitations of current technologies and highlight future research directions towards developing a multimedia conversational system.

We demonstrate ViDA-MAN, a digital-human agent for multi-modal interaction, which offers realtime audio-visual responses to instant speech inquiries. Compared to traditional text or voice-based system, ViDA-MAN offers human-like interactions (e.g, vivid voice, natural facial expression and body gestures). Given a speech request, the demonstration is able to response with high quality videos in sub-second latency. To deliver immersive user experience, ViDA-MAN seamlessly integrates multi-modal techniques including Acoustic Speech Recognition (ASR), multi-turn dialog, Text To Speech (TTS), talking heads video generation. Backed with large knowledge base, ViDA-MAN is able to chat with users on a number of topics including chit-chat, weather, device control, News recommendations, booking hotels, as well as answering questions via structured knowledge.

Due to the remarkable progress in recent years, deep face recognition is in great need of public support for practical model production and further exploration. The demands are in three folds, including 1) modular training scheme, 2) standard and automatic evaluation, and 3) groundwork of deployment. To meet these demands, we present a novel open-source project, named FaceX-Zoo, which is constructed with modular and scalable design, and oriented to the academic and industrial community of face-related analysis. FaceX-Zoo provides 1) the training module with various choices of backbone and supervisory head; 2) the evaluation module that enables standard and automatic test on most popular benchmarks; 3) the module of simple yet fully functional face SDK for the validation and primary application of end-to-end face recognition; 4) the additional module that integrates a group of useful tools. Based on these easy-to-use modules, FaceX-Zoo can help the community to easily build stateof-the-art solutions for deep face recognition and, such like the newly-emerged challenge of masked face recognition caused by the worldwide COVID-19 pandemic. Besides, FaceX-Zoo can be easily upgraded and scaled up along with further exploration in face related fields. The source codes and models have been released and received over 900 stars at https://github.com/JDAI-CV/FaceX-Zoo.

Digital image manipulations have been heavily abused to spread misinformation. Despite the great efforts dedicated in research community, prior works are mostly performance-driven, i.e., optimizing performances using standard/heavy networks designed for semantic classification. A thorough understanding for fake images detection models is still missing. This paper studies the essential ingredients for a good fake image detection model, by profiling the best-performing architectures. Specifically, we conduct a thorough analysis on a massive number of detection models, and observe how the performances are affected by different patterns of network structure. Our key findings include: 1) with the same computational budget, flat network structures (e.g., large kernel sizes, wide connections) perform better than commonly used deep networks; 2) operations in shallow layers deserve more computational capacities to trade-off performance and computational cost. These findings sketch a general profile for essential models of fake image detection, which show clear differences with those for semantic classification. Furthermore, based on our analysis, we propose a new Depth-Separable Search Space (DSS) for fake image detection. Compared to state-of-the-art methods, our model achieves competitive performance while saving more than 50% parameters.

Our work reveals a structured shortcoming of the existing mainstream self-supervised learning methods. Whereas self-supervised learning frameworks usually take the prevailing perfect instance level invariance hypothesis for granted, we carefully investigate the pitfalls behind. Particularly, we argue that the existing augmentation pipeline for generating multiple positive views naturally introduces out-of-distribution (OOD) samples that undermine the learning of the downstream tasks. Generating diverse positive augmentations on the input does not always pay off in benefiting downstream tasks. To overcome this inherent deficiency, we introduce a lightweight latent variable model UOTA, targeting the view sampling issue for self-supervised learning. UOTA adaptively searches for the most important sampling region to produce views, and provides viable choice for outlier-robust self-supervised learning approaches. Our method directly generalizes to many mainstream self-supervised learning approaches, regardless of the loss's nature contrastive or not. We empirically show UOTA's advantage over the state-of-the-art self-supervised paradigms with evident margin, which well justifies the existence of the OOD sample issue embedded in the existing approaches. Especially, we theoretically prove that the merits of the proposal boil down to guaranteed estimator variance and bias reduction. Code is available: https://github.com/ssl-codelab/uota.

In this paper, we tackle the problem of pose-guided person image generation with unpaired data, which is a challenging problem due to non-rigid spatial deformation. Instead of learning a fixed mapping directly between human bodies as previous methods, we propose a new pathway to decompose a single fixed mapping into two subtasks, namely, semantic parsing transformation and appearance generation. First, to simplify the learning for non-rigid deformation, a semantic generative network is developed to transform semantic parsing maps between different poses. Second, guided by semantic parsing maps, we render the foreground and background image, respectively. A foreground generative network learns to synthesize semantic-aware textures, and another background generative network learns to predict missing background regions caused by pose changes. Third, we enable pseudo-label training with unpaired data, and demonstrate that end-to-end training of the overall network further refines the semantic map prediction and final results accordingly. Moreover, our method is generalizable to other person image generation tasks defined on semantic maps, e.g., clothing texture transfer, controlled image manipulation, and virtual try-on. Experimental results on DeepFashion and Market-1501 datasets demonstrate the superiority of our method, especially in keeping better body shapes and clothing attributes, as well as rendering structure-coherent backgrounds.

Near-infrared to visible (NIR-VIS) face recognition is the most common case in heterogeneous face recognition, which aims to match a pair of face images captured from two different modalities. Existing deep learning based methods have made remarkable progress in NIR-VIS face recognition, while it encounters certain newly-emerged difficulties during the pandemic of COVID-19, since people are supposed to wear facial masks to cut off the spread of the virus. We define this task as NIR-VIS masked face recognition, and find it problematic with the masked face in the NIR probe image. First, the lack of masked face data is a challenging issue for the network training. Second, most of the facial parts (cheeks, mouth, nose etc.) are fully occluded by the mask, which leads to a large amount of loss of information. Third, the domain gap still exists in the remaining facial parts. In such scenario, the existing methods suffer from significant performance degradation caused by the above issues. In this paper, we aim to address the challenge of NIR-VIS masked face recognition from the perspectives of training data and training method. Specifically, we propose a novel heterogeneous training method to maximize the mutual information shared by the face representation of two domains with the help of semi-siamese networks. In addition, a 3D face reconstruction based approach is employed to synthesize masked face from the existing NIR image. Resorting to these practices, our solution provides the domain-invariant face representation which is also robust to the mask occlusion. Extensive experiments on three NIR-VIS face datasets demonstrate the effectiveness and cross-dataset-generalization capacity of our method.

Image captioning, aiming at generating natural sentences to describe image contents, has received significant attention with remarkable improvements in recent advances. The problem nevertheless is not trivial for cross-modal training due to the two challenges: 1) image detectors often consider only salient areas in an image and seldom explore the rich background context; 2) the language model is highly vulnerable to small but intentional perturbation attacks. To alleviate these issues, we propose the Noise Augmented Double-stream Graph Convolutional Networks (NADGCN) that novelly exploits the additional background context and enhances the generalization of the language model. Technically, NADGCN capitalizes on grid-stream GCN as a supplementary to the region stream, following the recipe that a rescaled grid graph can encode the relationship across grid areas over the full image rather than salient areas only. Moreover, we devise a noise module and integrate into the double-stream GCN to augment the capability of the basic generator. Such noise module introduces adaptive noise into the Recurrent Neural Networks (RNN) and is learnt through regarding the module as an agent with a stochastic Gaussian policy in Reinforcement Learning (RL). Extensive experiments on MSCOCO validate the design of the grid-stream GCN and the noise agent, and our generator outperforms the comparative baselines clearly.

Deep supervised hashing has emerged as an influential solution to large-scale semantic image retrieval problems in computer vision. In the light of recent progress, image label is the common way to define whether two images belong to the same category, but it contains little supervised information. The one-hot label can't accurately define the similarity of two images, which is important for image retrieval. In this paper, we propose an effective method, Deep Transfer Hashing(DTH) which uses the knowledge from teacher model as the supervised information. Inspired by knowledge distillation for model compression and deep hashing for fast image retrieval, we transfer the knowledge from a complex convolutional neural network(teacher) to a small neural network(student) which is used for fast image retrieval. The distance of the knowledge from teacher model can indicate the similarity of images. By minimizing the hashing codes distribution between the hashing layers of teacher model and student model, we can improve the retrieval performance. And we also evaluate the performance of the compressed model at inference stage. We test our method on widely used datasets CIFAR-10 and NUS-WIDE and we compare our method with other state-of-the-art methods in image retrieval domain. The experimental results show that our method can improve the image retrieval baseline by a large margin and better than other methods.

The task of reidentifying groups of people under different camera views is an important yet less-studied problem. Group reidentification (Re-ID) is a very challenging task since it is not only adversely affected by common issues in traditional single-object Re-ID problems, such as viewpoint and human pose variations, but also suffers from changes in group layout and group membership. In this paper, we propose a novel concept of group granularity by characterizing a group image by multigrained objects: individual people and subgroups of two and three people within a group. To achieve robust group Re-ID, we first introduce multigrained representations which can be extracted via the development of two separate schemes, that is, one with handcrafted descriptors and another with deep neural networks. The proposed representation seeks to characterize both appearance and spatial relations of multigrained objects, and is further equipped with importance weights which capture variations in intragroup dynamics. Optimal group-wise matching is facilitated by a multiorder matching process which, in turn, dynamically updates the importance weights in iterative fashion. We evaluated three multicamera group datasets containing complex scenarios and large dynamics, with experimental results demonstrating the effectiveness of our approach.

Recent advances in video object detection have characterized the exploration of temporal coherence across frames to enhance object detector. Nevertheless, previous solutions either rely on additional inputs (e.g., optical flow) to guide feature aggregation, or complex post-processing to associate bounding boxes. In this paper, we introduce a simple but effective design that learns instance identifiers for instance association in a meta-learning paradigm, which requires no auxiliary inputs or post-processing. Specifically, we present Meta-Learnt Instance Identifier Networks (namely MINet) that novelly meta-learns instance identifiers to recognize identical instances across frames in a single forward-pass, leading to the robust online linking of instances. Technically, depending on the detection results of previous frames, we teach MINet to learn the weights of an instance identifier on the fly, which can be well applied to up-coming frames. Such meta-learning paradigm enables instance identifiers to be flexibly adapted to novel frames at inference. Furthermore, MINet writes/updates the detection results of previous instances into memory and reads from memory when performing inference to encourage temporal consistency for video object detection. Our MINet is appealing in the sense that it is pluggable to any object detection model. Extensive experiments on ImageNet VID dataset demonstrate the superiority of MINet. More remarkably, by integrating MINet into Faster R-CNN, we achieve 80.2% mAP on ImageNet VID dataset.

Face parsing infers a pixel-wise label to each facial component, which has drawn much attention recently. Previous methods have shown their success in face parsing, which however overlook the correlation among facial components. As a matter of fact, the component-wise relationship is a critical clue in discriminating ambiguous pixels in facial area. To address this issue, we propose adaptive graph representation learning and reasoning over facial components, aiming to learn representative vertices that describe each component, exploit the component-wise relationship and thereby produce accurate parsing results against ambiguity. In particular, we devise an adaptive and differentiable graph abstraction method to represent the components on a graph via pixel-to-vertex projection under the initial condition of a predicted parsing map, where pixel features within a certain facial region are aggregated onto a vertex. Further, we explicitly incorporate the image edge as a prior in the model, which helps to discriminate edge and non-edge pixels during the projection, thus leading to refined parsing results along the edges. Then, our model learns and reasons over the relations among components by propagating information across vertices on the graph. Finally, the refined vertex features are projected back to pixel grids for the prediction of the final parsing map. To train our model, we propose a discriminative loss to penalize small distances between vertices in the feature space, which leads to distinct vertices with strong semantics. Experimental results show the superior performance of the proposed model on multiple face parsing datasets, along with the validation on the human parsing task to demonstrate the generalizability of our model.

Multi-person pose tracking task aims to estimate and track person keypoints in videos. Most of the previous methods follow the general track-by-detection strategy that ignores the consistent pose information during the whole framework. Thus, they often suffer from missing detections or inaccurate human association in challenging scenes with motion blur or person occlusion. To handle those problems, we propose a pose-guided tracking-by-detection framework that fuses pose information into both video human detection and human association procedures. In the video human detection stage, we adopt the pose-guided person location prediction exploiting the temporal information to make up missing detections. Technically, pose heatmaps are utilized to cope with the person-specific intra-class distractors. Furthermore, in the human association stage, we propose an appearance discriminative model based on the hierarchical pose-guided graph convolutional networks (PoseGCN). The PoseGCN-based model exploits human structural relations to boost person representation. Extensive experiments show the superiority of our method on the challenging pose tracking benchmark. Our proposed method ranks first on the PoseTrack leaderboard.11http://posetrack.net/leaderboard.php till the submission date (22-Aug-2019) of this paper. Our code has been publicly available at https://github.com/human-centric982/PGPT.

Single shot detectors that are potentially faster and simpler than two-stage detectors tend to be more applicable to object detection in videos. Nevertheless, the extension of such object detectors from image to video is not trivial especially when appearance deterioration exists in videos, e.g., motion blur or occlusion. A valid question is how to explore temporal coherence across frames for boosting detection. In this paper, we propose to address the problem by enhancing per-frame features through aggregation of neighboring frames. Specifically, we present Single Shot Video Object Detector (SSVD) - a new architecture that novelly integrates feature aggregation into a one-stage detector for object detection in videos. Technically, SSVD takes Feature Pyramid Network (FPN) as backbone network to produce multi-scale features. Unlike the existing feature aggregation methods, SSVD, on one hand, estimates the motion and aggregates the nearby features along the motion path, and on the other, hallucinates features by directly sampling features from the adjacent frames in a two-stream structure. Extensive experiments are conducted on ImageNet VID dataset, and competitive results are reported when comparing to state-of-the-art approaches. More remarkably, for $448 \times 448$ input, SSVD achieves 79.2% mAP on ImageNet VID, by processing one frame in 85 ms on an Nvidia Titan X Pascal GPU. The code is available at https://github.com/ddjiajun/SSVD.

In daily life, human beings rely on hands and body parts to complete particular actions cooperatively. These selected body parts and their cooperative relationships are essential cues to distinguish these actions. However, most existing action recognition methods, which try to model the body appearance or spatial relations in skeleton sequences, often ignore the essential cooperation relationship among joints. Differently, in this paper, we propose a spatio-temporal hierarchical soft quantization method to extract the congenerous motion features, which reflect the cooperation relations among joints and body parts. Specifically, we design a hierarchical network with multiple soft quantization layers to extract congenerous features. The hierarchical network not only models the spatial hierarchy of skeleton structure for joint, part, and body, but also extracts the temporal hierarchy with sliding windows for frame, fragment, and sequence. Moreover, the features in each layer are visually explainable, which reflect the cooperation among body parts. The trainable parameters in the network are also significantly reduced, which reduces computational cost. Extensive experiments conducted on four benchmarks demonstrate that our method can provide competitive results compared with state-of-the-arts. The visualized congenerous features also validate that our approach can effectively perceive the essential cooperation relations.

One fundamental challenge of vehicle re-identification (re-id) is to learn robust and discriminative visual representation, given the significant intra-class vehicle variations across different camera views. As the existing vehicle datasets are limited in terms of training images and viewpoints, we propose to build a unique large-scale vehicle dataset (called VehicleNet) by harnessing four public vehicle datasets, and design a simple yet effective two-stage progressive approach to learning more robust visual representation from VehicleNet. The first stage of our approach is to learn the generic representation for all domains (i.e., source vehicle datasets) by training with the conventional classification loss. This stage relaxes the full alignment between the training and testing domains, as it is agnostic to the target vehicle domain. The second stage is to fine-tune the trained model purely based on the target vehicle set, by minimizing the distribution discrepancy between our VehicleNet and any target domain. We discuss our proposed multi-source dataset VehicleNet and evaluate the effectiveness of the two-stage progressive representation learning through extensive experiments. We achieve the state-of-art accuracy of$\text{86.07}\%$mAP on the private test set of AICity Challenge, and competitive results on two other public vehicle re-id datasets, i.e., VeRi-776 and VehicleID. We hope this new VehicleNet dataset and the learned robust representations can pave the way for vehicle re-id in the real-world environments.

This article proposes a virtual leader-based coordinated controller for the nonlinear multiple autonomous underwater vehicles (multi-AUVs) with the system uncertainties. To achieve the coordinated formation, a virtual AUV is set as the leader, while the desired command is designed using the relative position between each AUV and the virtual leader. The controller is designed based on the back-stepping scheme, and the online data-based learning scheme is used for uncertainty approximation. The highlight is that compared with previous learning methods which mostly focus on stability, the learning performance index is constructed using the collected online data in this article. The index is further used in the composite update law of the neural weights. The closed-loop system stability is analyzed via the Lyapunov approach. The simulation test on the five AUVs under fixed formation shows that the proposed method can achieve higher tracking performance with improved approximation accuracy.

Live video broadcasting normally requires a multitude of skills and expertise with domain knowledge to enable multi-camera productions. As the number of cameras keeps increasing, directing a live sports broadcast has now become more complicated and challenging than ever before. The broadcast directors need to be much more concentrated, responsive, and knowledgeable, during the production. To relieve the directors from their intensive efforts, we develop an innovative automated sports broadcast directing system, called Smart Director, which aims at mimicking the typical human-in-the-loop broadcasting process to automatically create near-professional broadcasting programs in real-time by using a set of advanced multi-view video analysis algorithms. Inspired by the so-called “three-event” construction of sports broadcast [ 14 ], we build our system with an event-driven pipeline consisting of three consecutive novel components: (1) the Multi-View Event Localization to detect events by modeling multi-view correlations, (2) the Multi-View Highlight Detection to rank camera views by the visual importance for view selection, and (3) the Auto-Broadcasting Scheduler to control the production of broadcasting videos. To our best knowledge, our system is the first end-to-end automated directing system for multi-camera sports broadcasting, completely driven by the semantic understanding of sports events. It is also the first system to solve the novel problem of multi-view joint event detection by cross-view relation modeling. We conduct both objective and subjective evaluations on a real-world multi-camera soccer dataset, which demonstrate the quality of our auto-generated videos is comparable to that of the human-directed videos. Thanks to its faster response, our system is able to capture more fast-passing and short-duration events which are usually missed by human directors.

Face parsing has recently attracted increasing interest due to its numerous application potentials, such as facial make up and facial image generation. In this paper, we make contributions on face parsing task from two aspects. First, we develop a high-efficiency framework for pixel-level face parsing annotating and construct a new large-scale Landmark guided face Parsing dataset (LaPa). It consists of more than 22,000 facial images with abundant variations in expression, pose and occlusion, and each image of LaPa is provided with an 11-category pixel-level label map and 106-point landmarks. The dataset is publicly accessible to the community for boosting the advance of face parsing.1 Second, a simple yet effective Boundary-Attention Semantic Segmentation (BASS) method is proposed for face parsing, which contains a three-branch network with elaborately developed loss functions to fully exploit the boundary information. Extensive experiments on our LaPa benchmark and the public Helen dataset show the superiority of our proposed method.

Face recognition has witnessed significant progress due to the advances of deep convolutional neural networks (CNNs), the central task of which is how to improve the feature discrimination. To this end, several margin-based (e.g., angular, additive and additive angular margins) softmax loss functions have been proposed to increase the feature margin between different classes. However, despite great achievements have been made, they mainly suffer from three issues: 1) Obviously, they ignore the importance of informative features mining for discriminative learning; 2) They encourage the feature margin only from the ground truth class, without realizing the discriminability from other non-ground truth classes; 3) The feature margin between different classes is set to be same and fixed, which may not adapt the situations very well. To cope with these issues, this paper develops a novel loss function, which adaptively emphasizes the mis-classified feature vectors to guide the discriminative feature learning. Thus we can address all the above issues and achieve more discriminative face features. To the best of our knowledge, this is the first attempt to inherit the advantages of feature margin and feature mining into a unified loss function. Experimental results on several benchmarks have demonstrated the effectiveness of our method over state-of-the-art alternatives. Our code is available at http://www.cbsr.ia.ac.cn/users/xiaobowang/.

Region sampling or weighting is significantly important to the success of modern region-based object detectors. Unlike some previous works, which only focus on "hard'' samples when optimizing the objective function, we argue that sample weighting should be data-dependent and task-dependent. The importance of a sample for the objective function optimization is determined by its uncertainties to both object classification and bounding box regression tasks. To this end, we devise a general loss function to cover most region-based object detectors with various sampling strategies, and then based on it we propose a unified sample weighting network to predict a sample's task weights. Our framework is simple yet effective. It leverages the samples' uncertainty distributions on classification loss, regression loss, IoU, and probability score, to predict sample weights. Our approach has several advantages: (i). It jointly learns sample weights for both classification and regression tasks, which differentiates it from most previous work. (ii). It is a data-driven process, so it avoids some manual parameter tuning. (iii). It can be effortlessly plugged into most object detectors and achieves noticeable performance improvements without affecting their inference time. Our approach has been thoroughly evaluated with recent object detection frameworks and it can consistently boost the detection accuracy. Code has been made available at https://github.com/caiqi/sample-weighting-network.

Recent progress on fine-grained visual recognition and visual question answering has featured Bilinear Pooling, which effectively models the 2nd order interactions across multi-modal inputs. Nevertheless, there has not been evidence in support of building such interactions concurrently with attention mechanism for image captioning. In this paper, we introduce a unified attention block - X-Linear attention block, that fully employs bilinear pooling to selectively capitalize on visual information or perform multi-modal reasoning. Technically, X-Linear attention block simultaneously exploits both the spatial and channel-wise bilinear attention distributions to capture the 2ndorder interactions between the input single-modal or multi-modal features. Higher and even infinity order feature interactions are readily modeled through stacking multiple X-Linear attention blocks and equipping the block with Exponential Linear Unit (ELU) in a parameter-free fashion, respectively. Furthermore, we present X-Linear Attention Networks (dubbed as X-LAN) that novelly integrates X-Linear attention block(s) into image encoder and sentence decoder of image captioning model to leverage higher order intra- and inter-modal interactions. The experiments on COCO benchmark demonstrate that our X-LAN obtains to-date the best published CIDEr performance of 132.0% on COCO Karpathy test split. When further endowing Transformer with X-Linear attention blocks, CIDEr is boosted up to 132.8%. Source code is available at https://github.com/Panda-Peter/image-captioning.

Unsupervised domain adaptation has received significant attention in recent years. Most of existing works tackle the closed-set scenario, assuming that the source and target domains share the exactly same categories. In practice, nevertheless, a target domain often contains samples of classes unseen in source domain (i.e., unknown class). The extension of domain adaptation from closed-set to such open-set situation is not trivial since the target samples in unknown class are not expected to align with the source. In this paper, we address this problem by augmenting the state-of-the-art domain adaptation technique, Self-Ensembling, with category-agnostic clusters in target domain. Specifically, we present Self-Ensembling with Category-agnostic Clusters (SE-CC) --- a novel architecture that steers domain adaptation with the additional guidance of category-agnostic clusters that are specific to target domain. These clustering information provides domain-specific visual cues, facilitating the generalization of Self-Ensembling for both closed-set and open-set scenarios. Technically, clustering is firstly performed over all the unlabeled target samples to obtain the category-agnostic clusters, which reveal the underlying data space structure peculiar to target domain. A clustering branch is capitalized on to ensure that the learnt representation preserves such underlying structure by matching the estimated assignment distribution over clusters to the inherent cluster distribution for each target sample. Furthermore, SE-CC enhances the learnt representation with mutual information maximization. Extensive experiments are conducted on Office and VisDA datasets for both open-set and closed-set domain adaptation, and superior results are reported when comparing to the state-of-the-art approaches.

Semantic segmentation often requires a large set of images with pixel-level annotations. In the view of extremely expensive expert labeling, recent research has shown that the models trained on photo-realistic synthetic data (e.g., computer games) with computer-generated annotations can be adapted to real images. Despite this progress, without constraining the prediction on real images, the models will easily overfit on synthetic data due to severe domain mismatch. In this paper, we novelly exploit the intrinsic properties of semantic segmentation to alleviate such problem for model transfer. Specifically, we present a Regularizer of Prediction Transfer (RPT) that imposes the intrinsic properties as constraints to regularize model transfer in an unsupervised fashion. These constraints include patch-level, cluster-level and context-level semantic prediction consistencies at different levels of image formation. As the transfer is label-free and data-driven, the robustness of prediction is addressed by selectively involving a subset of image regions for model regularization. Extensive experiments are conducted to verify the proposal of RPT on the transfer of models trained on GTA5 and SYNTHIA (synthetic data) to Cityscapes dataset (urban street scenes). RPT shows consistent improvements when injecting the constraints on several neural networks for semantic segmentation. More remarkably, when integrating RPT into the adversarial-based segmentation framework, we report to-date the best results: mIoU of 53.2%/51.7% when transferring from GTA5/SYNTHIA to Cityscapes, respectively.

Most object recognition approaches predominantly focus on learning discriminative visual patterns, while overlooking the holistic object structure. Though important, structure modeling usually requires significant manual annotations and therefore is labor-intensive. In this paper, we propose to ``look into object" (explicitly yet intrinsically model the object structure) through incorporating self-supervisions into the traditional framework. We show the recognition backbone can be substantially enhanced for more robust representation learning, without any cost of extra annotation and inference speed. Specifically, we first propose an object-extent learning module for localizing the object according to the visual patterns shared among the instances in the same category. We then design a spatial context learning module for modeling the internal structures of the object, through predicting the relative positions within the extent. These two modules can be easily plugged into any backbone networks during training and detached at inference time. Extensive experiments show that our look-into-object approach (LIO) achieves large performance gain on a number of benchmarks, including generic object recognition (ImageNet) and fine-grained object recognition tasks (CUB, Cars, Aircraft). We also show that this learning paradigm is highly generalizable to other tasks such as object detection and segmentation (MS COCO). Project page: https://github.com/JDAI-CV/LIO.

In face recognition, designing margin-based (\emph{e.g.}, angular, additive, additive angular margins) softmax loss functions plays an important role to learn discriminative features. However, these hand-crafted heuristic methods may be sub-optimal because they require much effort to explore the large design space. Recently, an AutoML for loss function search method AM-LFS has been derived, which leverages reinforcement learning to search loss functions during the training process. But its search space is complex and unstable that hindering its superiority. In this paper, we first analyze that the key to enhance the feature discrimination is actually \textbf{how to reduce the softmax probability}. We then design a unified formulation for the current margin-based softmax losses. Accordingly, we define a novel search space and develop a reward-guided search method to automatically obtain the best candidate. Experimental results on a variety of face recognition benchmarks have demonstrated the effectiveness of our method over the state-of-the-art alternatives.

The current deep learning based visual tracking approaches have been very successful by learning the target classification and/or estimation model from a large amount of supervised training data in offline mode. However, most of them can still fail in tracking objects due to some more challenging issues such as dense distractor objects, confusing background, motion blurs, and so on. Inspired by the human “visual tracking” capability which leverages motion cues to distinguish the target from the background, we propose a Two-Stream Residual Convolutional Network (TS-RCN) for visual tracking, which successfully exploits both appearance and motion features for model update. Our TS-RCN can be integrated with existing deep learning based visual trackers. To further improve the tracking performance, we adopt a “wider” residual network ResNeXt as its feature extraction backbone. To the best of our knowledge, TS-RCN is the first end-to-end trainable two-stream visual tracking system, which makes full use of both appearance and motion features of the target. We have extensively evaluated the TS-RCN on most widely used benchmark datasets including VOT2018, VOT2019, and GOT-10K. The experiment results have successfully demonstrated that our two-stream model can greatly outperform the appearance-based tracker, and achieves state-of-the-art performance. The tracking system can run at up to 38.1 FPS.

Complementary recommendations, which aim at providing users product suggestions that are supplementary and compatible with their obtained items, have become a hot topic in both academia and industry in recent years. Existing work mainly focused on modeling the co-purchased relations between two items, but the compositional associations of item collections are largely unexplored. Actually, when a user chooses the complementary items for the purchased products, it is intuitive that she will consider the visual semantic coherence (such as color collocations, texture compatibilities) in addition to global impressions. Towards this end, in this paper, we propose a novel Content Attentive Neural Network (CANN) to model the comprehensive compositional coherence on both global contents and semantic contents. Specifically, we first propose a Global Coherence Learning (GCL) module based on multi-heads attention to model the global compositional coherence. Then, we generate the semantic-focal representations from different semantic regions and design a Focal Coherence Learning (FCL) module to learn the focal compositional coherence from different semantic-focal representations. Finally, we optimize the CANN in a novel compositional optimization strategy. Extensive experiments on the large-scale real-world data clearly demonstrate the effectiveness of CANN compared with several state-of-the-art methods.

Multi-person pose estimation has achieved great progress in recent years, even though, the precise prediction for occluded and invisible hard keypoints remains challenging. Most of the human pose estimation networks are equipped with an image classification-based pose encoder for feature extraction and a handcrafted pose decoder for high-resolution representations. However, the pose encoder might be sub-optimal because of the gap between image classification and pose estimation. The widely used multi-scale feature fusion in pose decoder is still coarse and cannot provide sufficient high-resolution details for hard keypoints. Neural Architecture Search (NAS) has shown great potential in many visual tasks to automatically search efficient networks. In this work, we present the Pose-native Network Architecture Search (PoseNAS) to simultaneously design a better pose encoder and pose decoder for pose estimation. Specifically, we directly search a data-oriented pose encoder with stacked searchable cells, which can provide an optimum feature extractor for the pose specific task. In the pose decoder, we exploit scale-adaptive fusion cells to promote rich information exchange across the multi-scale feature maps. Meanwhile, the pose decoder adopts a Fusion-and-Enhancement manner to progressively boost the high-resolution representations that are non-trivial for the precious prediction of hard keypoints. With the exquisitely designed search space and search strategy, PoseNAS can simultaneously search all modules in an end-to-end manner. PoseNAS achieves state-of-the-art performance on three public datasets, MPII, COCO, and PoseTrack, with small-scale parameters compared with the existing methods. Our best model obtains 76.7% mAP and 75.9% mAP on the COCO validation set and test set with only 33.6M parameters. Code and implementation are available at https://github.com/for-code0216/PoseNAS.

This demonstration presents an instant and progressive cross-modality person search system, called 'CMPS'. Through the system, users can instantly find the lost children or elderly persons by simply describing their appearance through speech. Unlike most existing person search applications which have to cost much time to find the probe images, CMPS will save more valuable time in the early stage of losing. The proposed CMPS is one of the first attempts towards instant and progressive person search leveraging the audio, text, and visual modalities together. In detail, the system first takes the speech that describes the appearance of a person as the input to obtain a textual description by speech-to-text conversion. Then the cross-modal search is performed by matching the textual embedding with the visual representations of images in the learned latent space. The searched images can be used as candidates for query expansion. If the candidates are not right, the user can quickly adjust their description through speech. Once a right image is found, the user can directly click it as a new query. Finally the system will give the complete track of the lost person by once-click. On the built CUHK-PEDES-AUDIOS dataset, the system can achieve 82.46% rank-1 accuracy in real-time speed. Our code of CMPS is available at https://github.com/SheldongChen/Search-People-With-Audio.

Video anomaly detection is an essential task in computer vision which attracts massive attention from academia and industry. The existing approaches are implemented in diverse deep learning frameworks and settings, making it difficult to reproduce the results published by the original authors. Undoubtedly, this phenomenon is detrimental to the development of Video Anomaly detection and community communication. In this paper, we present a PyTorch-based video anomaly detection toolbox, namely PyAnomaly that contains high modular and extensible components, comprehensive and impartial evaluation platforms, a friendly manageable system configuration, and the abundant engineering deployment functions. To make it easy-to-use and easy-to-extend, we implement the architecture by hooks and registers functionality. Remarkably, we have reproduced the comparable experimental results of six representative methods as those published by the original authors, and we will release these pre-trained models with more rich configurations. To our best knowledge, the PyAnomaly is the first open-source tool in video anomaly detection and is available at https://github.com/YuhaoCheng/PyAnomaly.

Human free-hand sketches have been studied in various fields including sketch recognition, synthesis and sketch-based image retrieval. We propose a new challenging task sketch enhancement (SE) defined in an ill-posed space, i.e. enhancing a non-professional sketch (NPS) to a professional sketch (PS), which is a creative generation task different from sketch abstraction, sketch completion and sketch variation. For the first time we release a database of NPS with PS for anime characters. We cast sketch enhancement as an image-to-image translation problem by exploiting the relationship to corresponding intensive or sparse pixel domains for sketch domain. Specifically, we explore three different routines based on conditional generative adversarial network (cGAN), i.e. Sketch-Sketch (SS), Sketch-Colorization-Sketch (SCS) and Sketch-Abstraction-Sketch (SAS). SS is a one-stage model that directly maps NPS to PS, while SCS and SAS are two-stage models where auxiliary inputs, grayscale parsing and shape parsing, are involved. Multiple metrics are used to evaluate the performance of the models in both the sketch domain and other low-level feature domains. With quantitative and qualitative analysis of the experiments, we have established solid baselines, which, we hope, could encourage more research conducted on this task. Our dataset is publicly available via https://github.com/LCXCUC/SketchMan2020.

Vehicle re-identification (Re-Id) is a challenging task due to the inter-class similarity, the intra-class difference, and the cross-view misalignment of vehicle parts. Although recent methods achieve great improvement by learning detailed features from keypoints or bounding boxes of parts, vehicle Re-Id is still far from being solved. Different from existing methods, we propose a Parsing-guided Cross-part Reasoning Network, named as PCRNet, for vehicle Re-Id. The PCRNet explores vehicle parsing to learn discriminative part-level features, model the correlation among vehicle parts, and achieve precise part alignment for vehicle Re-Id. To accurately segment vehicle parts, we first build a large-scale Multi-grained Vehicle Parsing (MVP) dataset from surveillance images. With the parsed parts, we extract regional features for each part and build a part-neighboring graph to explicitly model the correlation among parts. Then, the graph convolutional networks (GCNs) are adopted to propagate local information among parts, which can discover the most effective local features of varied viewpoints. Moreover, we propose a self-supervised part prediction loss to make the GCNs generate features of invisible parts from visible parts under different viewpoints. By this means, the same vehicle from different viewpoints can be matched with the well-aligned and robust feature representations. Through extensive experiments, our PCRNet significantly outperforms the state-of-the-art methods on three large-scale vehicle Re-Id datasets.

Virtual try-on has attracted lots of research attention due to its potential applications in e-commerce, virtual reality and fashion design. However, existing methods can hardly preserve the fine-grained details (e.g., clothing texture, facial identity, hair style, skin tone) during generation, due to the non-rigid body deformation and multi-scale details. In this work, we propose a multi-stage framework to synthesize person images, where fine-grained details can be well preserved. To address the long-range translation and rich-details generation, we propose a Tree-Block (tree dilated fusion block) to replace standard ResNet-block where applicable. Notably, multi-scale feature maps can be smoothly fused for fine-grained detail generation, by incorporating larger spatial context at multiple scales. With a delicate end-to-end training scheme, our whole framework can be jointly optimized for results with significantly better visual fidelity and richer details. Moreover, we also explore the potential application in video-based virtual try-on. By harnessing the well-trained image generator and an extra video-level adaptor, a model photo can be well animated with a driving pose sequence. Extensive evaluations on standard datasets and user study demonstrate that our proposed framework achieves the state-of-the-art results, especially in preserving visual details in clothing texture and facial identity. Our implementation is publicly available via https://github.com/JDAI-CV/Down-to-the-Last-Detail-Virtual-Try-on-with-Detail-Carving.

Person re-identification (Re-ID) aims at retrieving an input person image from a set of images captured by multiple cameras. Although recent Re-ID methods have made great success, most of them extract features in terms of the attributes of clothing (e.g., color, texture). However, it is common for people to wear black clothes or be captured by surveillance systems in low light illumination, in which cases the attributes of the clothing are severely missing. We call this problem the Black Re-ID problem. To solve this problem, rather than relying on the clothing information, we propose to exploit head-shoulder features to assist person Re-ID. The head-shoulder adaptive attention network (HAA) is proposed to learn the head-shoulder feature and an innovative ensemble method is designed to enhance the generalization of our model. Given the input person image, the ensemble method would focus on the head-shoulder feature by assigning a larger weight if the individual insides the image is in black clothing. Due to the lack of a suitable benchmark dataset for studying the Black Re-ID problem, we also contribute the first Black-reID dataset, which contains 1274 identities in training set. Extensive evaluations on the Black-reID, Market1501 and DukeMTMC-reID datasets show that our model achieves the best result compared with the state-of-the-art Re-ID methods on both Black and conventional Re-ID problems. Furthermore, our method is also proved to be effective in dealing with person Re-ID in similar clothing. Our code and dataset are avaliable on https://github.com/xbq1994/.

Real-time in-match soccer statistics provide continuous tracking of soccer ball and player positions and speeds, enabling advanced analytics. Currently, only elite soccer leagues have the luxury of tracking in-match soccer statistics operated with a large number of trained personnel. In this work, we present an Automated In-match Soccer Analysis System (AI-SAS), using a domain-knowledge-based multi-view global tracking. This system tracks player team, position, and speed automatically, providing real-time in-match team- and individual-level statistics and analyses. In comparison with the latest soccer analysis systems, AI-SAS is more scalable in streaming multiple video sources for real-time process and more flexible in hosting plug-and-play deep-learning-based tracking-by-detection algorithms. The global multi-view tracking also overcomes the single-view limitation and improves the tracking accuracy.

Text-based person search aims to retrieve the pedestrian images that best match a given textual description from gallery images. Previous methods utilize the soft-attention mechanism to infer the semantic alignments between the regions of image and the corresponding words in sentence. However, these methods may fuse the irrelevant multi-modality features together which cause matching redundancy problem. In this work, we propose a novel hierarchical Gumbel attention network for text-based person search via Gumbel top-k re-parameterization algorithm. Specifically, it adaptively selects the strong semantically relevant image regions and words/phrases from images and texts for precise alignment and similarity calculation. This hard selection strategy is able to fuse the strong-relevant multi-modality features for alleviating the problem of matching redundancy. Meanwhile, a Gumbel top-k re-parameterization algorithm is designed as a low-variance, unbiased gradient estimator to handle the discreteness problem of hard attention mechanism by an end-to-end manner. Moreover, a hierarchical adaptive matching strategy is employed by the model from three different granularities, i.e., word-level, phrase-level, and sentence-level, towards fine-grained matching. Extensive experimental results demonstrate the state-of-the-art performance. Compared the existed best method, we achieve the 8.24% Rank-1 and 7.6% mAP relative improvements in the text-to-image retrieval task, and 5.58% Rank-1 and 6.3% mAP relative improvements in the image-to-text retrieval task on CUHK-PEDES dataset, respectively.

Live sports broadcasting is the live coverage of sports (e.g., a soccer match) as a television program, on various types of broadcasting media (e.g., television or internet). Directing such live sports broadcast is cost-expensive and demands experienced sports directors with sufficient broadcasting skills. In this paper, we demonstrate an end-to-end intelligent system for live sports broadcasting, namely iDirector, which aims to mimic the human-in-loop live broadcasting process by aggregating the input multi-camera video streaming into the final output program video (PGM video) for audience. We construct this system as an event-driven pipeline with three modules: video decoder, video analyzer, and broadcasting controller. Specifically, given the multi-view video streaming captured from cameras placing in the stadium, video decoder module first decodes the input video streaming into a series of frames and clips. Next, video analyzer performs multiple pre-learned models in parallel for frame- and clip-level content understanding (e.g., events localization and highlight detection). Based on all the analytic results across frames and clips, with only 30 seconds looking ahead, broadcasting controller automatically produces the broadcast videos via camera view switch, playback and slow-motion. When some high-profile events (e.g., free kick) happen, broadcasting controller will render visual effects on PGM video to enhance audiences' entertained pleasure.

This paper explores useful modifications of the recent development in contrastive learning via novel probabilistic modeling. We derive a particular form of contrastive loss named Joint Contrastive Learning (JCL). JCL implicitly involves the simultaneous learning of an infinite number of query-key pairs, which poses tighter constraints when searching for invariant features. We derive an upper bound on this formulation that allows analytical solutions in an end-to-end training manner. While JCL is practically effective in numerous computer vision applications, we also theoretically unveil the certain mechanisms that govern the behavior of JCL. We demonstrate that the proposed formulation harbors an innate agency that strongly favors similarity within each instance-specific class, and therefore remains advantageous when searching for discriminative features among distinct instances. We evaluate these proposals on multiple benchmarks, demonstrating considerable improvements over existing algorithms. Code is publicly available at: https://github.com/caiqi/Joint-Contrastive-Learning.

With the advancement of image processing and computer vision technology, content-based product search is applied in a wide variety of common tasks, such as online shopping, automatic checkout systems, and intelligent logistics. Given a product image as a query, existing product search systems mainly perform the retrieval process using predefined databases with fixed product categories. However, real-world applications often require inserting new categories or updating existing products in the product database. When using existing product search methods, the image feature extraction models must be retrained and database indexes must be rebuilt to accommodate the updated data, and these operations incur high costs for data annotation and training time. To this end, we propose a few-shot incremental product search framework with meta-learning, which requires very few annotated images and has a reasonable training time. In particular, our framework contains a multipooling-based product feature extractor that learns a discriminative representation for each product, and we also design a meta-learning-based feature adapter to guarantee the robustness of the few-shot features. Furthermore, when expanding new categories in batches during a product search, we reconstruct the few-shot features by using an incremental weight combiner to accommodate the incremental search task. Through extensive experiments, we demonstrate that the proposed framework achieves excellent performance for new products while still guaranteeing the high search accuracy of the base categories after gradually expanding new product categories without forgetting.

We investigate the localization of subtle yet discriminative parts for fine-grained image recognition. Based on the observation that such parts typically exist within a hierarchical structure (e.g., from a coarse-scale "head" to a fine-scale "eye" when recognizing bird species), we propose a novel progressive-attention convolutional neural network (PA-CNN) to progressively localize parts at multiple scales. The PA-CNN localizes parts in two steps, where a part proposal network (PPN) generates multiple local attention maps, and a part rectification network (PRN) learns part-specific features from each proposal and provides the PPN with refined part locations. This coupling of the PPN and PRN allows them to be optimized in a mutually reinforcing manner, leading to improved pinpointing of fine-grained parts. Moreover, the convolutional parameters for a PPN at a finer scale can be inherited from the PRN at a coarser scale, enabling a rich part hierarchy (e.g., eye and beak in a bird's head) to be learned in a stacked fashion. Case studies show that PA-CNN can precisely identify parts without using bounding box/part annotations. In addition, quantitative evaluations demonstrate that PA-CNN yields state-of-the-art performance in three challenging fine-grained recognition tasks. i.e., CUB-200-2011, FGVC-Aircraft, and Stanford Cars.

Retweeting is a powerful driving force in information propagation on microblogging sites. However, identifying the most effective retweeters of a message (called the ”key retweeter prediction” problem) has become a significant research topic. Conventional approaches have addressed this topic from two main aspects: by analyzing either the personal attributes of microblogging users or the structures of user graph networks. However, according to sociological findings, author-retweeter dependencies also play a crucial role in influencing message propagation. In this paper, we propose a novel model to solve the key retweeter prediction problem by incorporating the auxiliary relations between a tweet author and potential retweeters. Without loss of generality, we formulate the relations from four relational factors: status relation, temporal relation, locational relation, and interactive relation. In addition, we propose a novel method, called “Relation-based Learning to Rank (RL2R),” to determine the key retweeters for a given tweet by ranking the potential retweeters in terms of their spreadability. The experimental results show that our method outperforms the state-of-the-art algorithms at top-k retweeter prediction, achieving a significant relative average improvement of 19.7-29.4 percent. These findings provide new insights for understanding user behaviors on social media for key retweeter prediction purposes.

The problem of distance metric learning is mostly considered from the perspective of learning an embedding space, where the distances between pairs of examples are in correspondence with a similarity metric. With the rise and success of Convolutional Neural Networks (CNN), deep metric learning (DML) involves training a network to learn a nonlinear transformation to the embedding space. Existing DML approaches often express the supervision through maximizing inter-class distance and minimizing intra-class variation. However, the results can suffer from overfitting problem, especially when the training examples of each class are embedded together tightly and the density of each class is very high. In this paper, we integrate density, i.e., the measure of data concentration in the representation, into the optimization of DML frameworks to adaptively balance inter-class similarity and intra-class variation by training the architecture in an end-to-end manner. Technically, the knowledge of density is employed as a regularizer, which is pluggable to any DML architecture with different objective functions such as contrastive loss, N-pair loss and triplet loss. Extensive experiments on three public datasets consistently demonstrate clear improvements by amending three types of embedding with the density adaptivity. More remarkably, our proposal increases Recall@1 from 67.95% to 77.62%, from 52.01% to 55.64% and from 68.20% to 70.56% on Cars196, CUB-200-2011 and Stanford Online Products dataset, respectively.

Localizing temporal action proposals from long videos is a fundamental challenge in video analysis (e.g., action detection and recognition or dense video captioning). Most existing approaches often overlook the hierarchical granularities of actions and thus fail to discriminate fine-grained action proposals (e.g., hand washing laundry or changing a tire in vehicle repair). In this paper, we propose a novel coarse-to-fine temporal proposal (CFTP) approach to localize temporal action proposals by exploring different action granularities. Our proposed CFTP consists of three stages: a coarse proposal network (CPN) to generate long action proposals, a temporal convolutional anchor network (CAN) to localize finer proposals, and a proposal reranking network (PRN) to further identify proposals from previous stages. Specifically, CPN explores three complementary actionness curves (namely pointwise, pairwise, and recurrent curves) that represent actions at different levels for generating coarse proposals, while CAN refines these proposals by a multiscale cascaded 1D-convolutional anchor network. In contrast to existing works, our coarse-to-fine approach can progressively localize fine-grained action proposals. We conduct extensive experiments on two action benchmarks (THUMOS14 and ActivityNet v1.3) and demonstrate the superior performance of our approach when compared to the state-of-the-art techniques on various video understanding tasks.

Person search with one portrait, which attempts to search the targets in arbitrary scenes using one portrait image at a time, is an essential yet unexplored problem in the multimedia field. Existing approaches, which predominantly depend on the visual information of persons, cannot solve problems when there are variations in the person’s appearance caused by complex environments and changes in pose, makeup, and clothing. In contrast to existing methods, in this article, we propose an associative multimodality index for person search with face, body, and voice information. In the offline stage, an associative network is proposed to learn the relationships among face, body, and voice information. It can adaptively estimate the weights of each embedding to construct an appropriate representation. The multimodality index can be built by using these representations, which exploit the face and voice as long-term keys and the body appearance as a short-term connection. In the online stage, through the multimodality association in the index, we can retrieve all targets depending only on the facial features of the query portrait. Furthermore, to evaluate our multimodality search framework and facilitate related research, we construct the Cast Search in Movies with Voice (CSM-V) dataset, a large-scale benchmark that contains 127K annotated voices corresponding to tracklets from 192 movies. According to extensive experiments on the CSM-V dataset, the proposed multimodality person search framework outperforms the state-of-the-art methods.

It is well believed that video captioning is a fundamental but challenging task in both computer vision and artificial intelligence fields. The prevalent approach is to map an input video to a variable-length output sentence in a sequence to sequence manner via Recurrent Neural Network (RNN). Nevertheless, the training of RNN still suffers to some degree from vanishing/exploding gradient problem, making the optimization difficult. Moreover, the inherently recurrent dependency in RNN prevents parallelization within a sequence during training and therefore limits the computations. In this paper, we present a novel design — Temporal Deformable Convolutional Encoder-Decoder Networks (dubbed as TDConvED) that fully employ convolutions in both encoder and decoder networks for video captioning. Technically, we exploit convolutional block structures that compute intermediate states of a fixed number of inputs and stack several blocks to capture long-term relationships. The structure in encoder is further equipped with temporal deformable convolution to enable free-form deformation of temporal sampling. Our model also capitalizes on temporal attention mechanism for sentence generation. Extensive experiments are conducted on both MSVD and MSR-VTT video captioning datasets, and superior results are reported when comparing to conventional RNN-based encoder-decoder techniques. More remarkably, TDConvED increases CIDEr-D performance from 58.8% to 67.2% on MSVD.

To date, visual question answering (VQA) (i.e., image QA and video QA) is still a holy grail in vision and language understanding, especially for video QA. Compared with image QA that focuses primarily on understanding the associations between image region-level details and corresponding questions, video QA requires a model to jointly reason across both spatial and long-range temporal structures of a video as well as text to provide an accurate answer. In this paper, we specifically tackle the problem of video QA by proposing a Structured Two-stream Attention network, namely STA, to answer a free-form or open-ended natural language question about the content of a given video. First, we infer rich longrange temporal structures in videos using our structured segment component and encode text features. Then, our structured two-stream attention component simultaneously localizes important visual instance, reduces the influence of background video and focuses on the relevant text. Finally, the structured two-stream fusion component incorporates different segments of query and video aware context representation and infers the answers. Experiments on the large-scale video QA dataset TGIF-QA show that our proposed method significantly surpasses the best counterpart (i.e., with one representation for the video input) by 13.0%, 13.5%, 11.0% and 0.3 for Action, Trans., TrameQA and Count tasks. It also outperforms the best competitor (i.e., with two representations) on the Action, Trans., TrameQA tasks by 4.1%, 4.7%, and 5.1%.

We have witnessed the tremendous growth of videos over the Internet, where most of these videos are typically paired with abundant sentence descriptions, such as video titles, captions and comments. Therefore, it has been increasingly crucial to associate specific video segments with the corresponding informative text descriptions, for a deeper understanding of video content. This motivates us to explore an overlooked problem in the research community — temporal sentence localization in video, which aims to automatically determine the start and end points of a given sentence within a paired video. For solving this problem, we face three critical challenges: (1) preserving the intrinsic temporal structure and global context of video to locate accurate positions over the entire video sequence; (2) fully exploring the sentence semantics to give clear guidance for localization; (3) ensuring the efficiency of the localization method to adapt to long videos. To address these issues, we propose a novel Attention Based Location Regression (ABLR) approach to localize sentence descriptions in videos in an efficient end-to-end manner. Specifically, to preserve the context information, ABLR first encodes both video and sentence via Bi-directional LSTM networks. Then, a multi-modal co-attention mechanism is presented to generate both video and sentence attentions. The former reflects the global video structure, while the latter highlights the sentence details for temporal localization. Finally, a novel attention based location prediction network is designed to regress the temporal coordinates of sentence from the previous attentions. We evaluate the proposed ABLR approach on two public datasets ActivityNet Captions and TACoS. Experimental results show that ABLR significantly outperforms the existing approaches in both effectiveness and efficiency.

Delicate feature representation about object parts plays a critical role in fine-grained recognition. For example, experts can even distinguish fine-grained objects relying only on object parts according to professional knowledge. In this paper, we propose a novel "Destruction and Construction Learning" (DCL) method to enhance the difficulty of fine-grained recognition and exercise the classification model to acquire expert knowledge. Besides the standard classification backbone network, another "destruction and construction" stream is introduced to carefully "destruct" and then "reconstruct" the input image, for learning discriminative regions and features. More specifically, for "destruction", we first partition the input image into local regions and then shuffle them by a Region Confusion Mechanism (RCM). To correctly recognize these destructed images, the classification network has to pay more attention to discriminative regions for spotting the differences. To compensate the noises introduced by RCM, an adversarial loss, which distinguishes original images from destructed ones, is applied to reject noisy patterns introduced by RCM. For "construction", a region alignment network, which tries to restore the original spatial layout of local regions, is followed to model the semantic correlation among local regions. By jointly training with parameter sharing, our proposed DCL injects more discriminative local details to the classification network. Experimental results show that our proposed framework achieves state-of-the-art performance on three standard benchmarks. Moreover, our proposed method does not need any external knowledge during training, and there is no computation overhead at inference time except the standard classification network feed-forwarding. Source code: https://github.com/JDAI-CV/DCL.

Image captioning has received significant attention with remarkable improvements in recent advances. Nevertheless, images in the wild encapsulate rich knowledge and cannot be sufficiently described with models built on image-caption pairs containing only in-domain objects. In this paper, we propose to address the problem by augmenting standard deep captioning architectures with object learners. Specifically, we present Long Short-Term Memory with Pointing (LSTM-P) --- a new architecture that facilitates vocabulary expansion and produces novel objects via pointing mechanism. Technically, object learners are initially pre-trained on available object recognition data. Pointing in LSTM-P then balances the probability between generating a word through LSTM and copying a word from the recognized objects at each time step in decoder stage. Furthermore, our captioning encourages global coverage of objects in the sentence. Extensive experiments are conducted on both held-out COCO image captioning and ImageNet datasets for describing novel objects, and superior results are reported when comparing to state-of-the-art approaches. More remarkably, we obtain an average of 60.9% in F1 score on held-out COCO dataset.

Discovering social relations, e.g., kinship, friendship, etc., from visual contents can make machines better interpret the behaviors and emotions of human beings. Existing studies mainly focus on recognizing social relations from still images while neglecting another important media--video. On one hand, the actions and storylines in videos provide more important cues for social relation recognition. On the other hand, the key persons may appear at arbitrary spatial-temporal locations, even not in one same image from beginning to the end. To overcome these challenges, we propose a Multi-scale Spatial-Temporal Reasoning (MSTR) framework to recognize social relations from videos. For the spatial representation, we not only adopt a temporal segment network to learn global action and scene information, but also design a Triple Graphs model to capture visual relations between persons and objects. For the temporal domain, we propose a Pyramid Graph Convolutional Network to perform temporal reasoning with multi-scale receptive fields, which can obtain both long-term and short-term storylines in videos. By this means, MSTR can comprehensively explore the multi-scale actions and storylines in spatial-temporal dimensions for social relation reasoning in videos. Extensive experiments on a new large-scale Video Social Relation dataset demonstrate the effectiveness of the proposed framework.

Temporally localizing actions in a video is a fundamental challenge in video understanding. Most existing approaches have often drawn inspiration from image object detection and extended the advances, e.g., SSD and Faster R-CNN, to produce temporal locations of an action in a 1D sequence. Nevertheless, the results can suffer from robustness problem due to the design of predetermined temporal scales, which overlooks the temporal structure of an action and limits the utility on detecting actions with complex variations. In this paper, we propose to address the problem by introducing Gaussian kernels to dynamically optimize temporal scale of each action proposal. Specifically, we present Gaussian Temporal Awareness Networks (GTAN) - a new architecture that novelly integrates the exploitation of temporal structure into an one-stage action localization framework. Technically, GTAN models the temporal structure through learning a set of Gaussian kernels, each for a cell in the feature maps. Each Gaussian kernel corresponds to a particular interval of an action proposal and a mixture of Gaussian kernels could further characterize action proposals with various length. Moreover, the values in each Gaussian curve reflect the contextual contributions to the localization of an action proposal. Extensive experiments are conducted on both THUMOS14 and ActivityNet v1.3 datasets, and superior results are reported when comparing to state-of-the-art approaches. More remarkably, GTAN achieves 1.9% and 1.1% improvements in mAP on testing set of the two datasets.

In this paper, we introduce a new idea for unsupervised domain adaptation via a remold of Prototypical Networks, which learn an embedding space and perform classification via a remold of the distances to the prototype of each class. Specifically, we present Transferrable Prototypical Networks (TPN) for adaptation such that the prototypes for each class in source and target domains are close in the embedding space and the score distributions predicted by prototypes separately on source and target data are similar. Technically, TPN initially matches each target example to the nearest prototype in the source domain and assigns an example a ``pseudo" label. The prototype of each class could then be computed on source-only, target-only and source-target data, respectively. The optimization of TPN is end-to-end trained by jointly minimizing the distance across the prototypes on three types of data and KL-divergence of score distributions output by each pair of the prototypes. Extensive experiments are conducted on the transfers across MNIST, USPS and SVHN datasets, and superior results are reported when comparing to state-of-the-art approaches. More remarkably, we obtain an accuracy of 80.4% of single model on VisDA 2017 dataset.

Convolutional Neural Networks (CNN) have been regarded as a powerful class of models for visual recognition problems. Nevertheless, the convolutional filters in these networks are local operations while ignoring the large-range dependency. Such drawback becomes even worse particularly for video recognition, since video is an information-intensive media with complex temporal variations. In this paper, we present a novel framework to boost the spatio-temporal representation learning by Local and Global Diffusion (LGD). Specifically, we construct a novel neural network architecture that learns the local and global representations in parallel. The architecture is composed of LGD blocks, where each block updates local and global features by modeling the diffusions between these two representations. Diffusions effectively interact two aspects of information, i.e., localized and holistic, for more powerful way of representation learning. Furthermore, a kernelized classifier is introduced to combine the representations from two aspects for video recognition. Our LGD networks achieve clear improvements on the large-scale Kinetics-400 and Kinetics-600 video classification datasets against the best competitors by 3.5% and 0.7%. We further examine the generalization of both the global and local representations produced by our pre-trained LGD networks on four different benchmarks for video action recognition and spatio-temporal action detection tasks. Superior performances over several state-of-the-art techniques on these benchmarks are reported.

In this paper, we address unsupervised pose-guided person image generation, which is known challenging due to non-rigid deformation. Unlike previous methods learning a rock-hard direct mapping between human bodies, we propose a new pathway to decompose the hard mapping into two more accessible subtasks, namely, semantic parsing transformation and appearance generation. Firstly, a semantic generative network is proposed to transform between semantic parsing maps, in order to simplify the non-rigid deformation learning. Secondly, an appearance generative network learns to synthesize semantic-aware textures. Thirdly, we demonstrate that training our framework in an end-to-end manner further refines the semantic maps and final results accordingly. Our method is generalizable to other semantic-aware person image generation tasks, e.g., clothing texture transfer and controlled image manipulation. Experimental results demonstrate the superiority of our method on DeepFashion and Market-1501 datasets, especially in keeping the clothing attributes and better body shapes.

In this paper, we propose a Customizable Architecture Search (CAS) approach to automatically generate a network architecture for semantic image segmentation. The generated network consists of a sequence of stacked computation cells. A computation cell is represented as a directed acyclic graph, in which each node is a hidden representation (i.e., feature map) and each edge is associated with an operation (e.g., convolution and pooling), which transforms data to a new layer. During the training, the CAS algorithm explores the search space for an optimized computation cell to build a network. The cells of the same type share one architecture but with different weights. In real applications, however, an optimization may need to be conducted under some constraints such as GPU time and model size. To this end, a cost corresponding to the constraint will be assigned to each operation. When an operation is selected during the search, its associated cost will be added to the objective. As a result, our CAS is able to search an optimized architecture with customized constraints. The approach has been thoroughly evaluated on Cityscapes and CamVid datasets, and demonstrates superior performance over several state-of-the-art techniques. More remarkably, our CAS achieves 72.3% mIoU on the Cityscapes dataset with speed of 108 FPS on an Nvidia TitanXp GPU.

Current state-of-the-art object objectors are fine-tuned from the off-the-shelf networks pretrained on large-scale classification dataset ImageNet, which incurs some additional problems: 1) The classification and detection have different degrees of sensitivity to translation, resulting in the learning objective bias; 2) The architecture is limited by the classification network, leading to the inconvenience of modification. To cope with these problems, training detectors from scratch is a feasible solution. However, the detectors trained from scratch generally perform worse than the pretrained ones, even suffer from the convergence issue in training. In this paper, we explore to train object detectors from scratch robustly. By analysing the previous work on optimization landscape, we find that one of the overlooked points in current trained-from-scratch detector is the BatchNorm. Resorting to the stable and predictable gradient brought by BatchNorm, detectors can be trained from scratch stably while keeping the favourable performance independent to the network architecture. Taking this advantage, we are able to explore various types of networks for object detection, without suffering from the poor convergence. By extensive experiments and analyses on downsampling factor, we propose the Root-ResNet backbone network, which makes full use of the information from original images. Our ScratchDet achieves the state-of-the-art accuracy on PASCAL VOC 2007, 2012 and MS COCO among all the train-from-scratch detectors and even performs better than several one-stage pretrained methods. Codes will be made publicly available at https://github.com/KimSoybean/ScratchDet.

It has been well recognized that modeling object-to-object relations would be helpful for object detection. Nevertheless, the problem is not trivial especially when exploring the interactions between objects to boost video object detectors. The difficulty originates from the aspect that reliable object relations in a video should depend on not only the objects in the present frame but also all the supportive objects extracted over a long range span of the video. In this paper, we introduce a new design to capture the interactions across the objects in spatio-temporal context. Specifically, we present Relation Distillation Networks (RDN) - a new architecture that novelly aggregates and propagates object relation to augment object features for detection. Technically, object proposals are first generated via Region Proposal Networks (RPN). RDN then, on one hand, models object relation via multi-stage reasoning, and on the other, progressively distills relation through refining supportive object proposals with high objectness scores in a cascaded manner. The learnt relation verifies the efficacy on both improving object detection in each frame and box linking across frames. Extensive experiments are conducted on ImageNet VID dataset, and superior results are reported when comparing to state-of-the-art methods. More remarkably, our RDN achieves 81.8% and 83.2% mAP with ResNet-101 and ResNeXt-101, respectively. When further equipped with linking and rescoring, we obtain to-date the best reported mAP of 83.8% and 84.7%.

Relationships encode the interactions among individual instances and play a critical role in deep visual scene understanding. Suffering from the high predictability with non-visual information, relationship models tend to fit the statistical bias rather than ``learning" to infer the relationships from images. To encourage further development in visual relationships, we propose a novel method to mine more valuable relationships by automatically pruning visually-irrelevant relationships. We construct a new scene graph dataset named Visually-Relevant Relationships Dataset (VrR-VG) based on Visual Genome. Compared with existing datasets, the performance gap between learnable and statistical method is more significant in VrR-VG, and frequency-based analysis does not work anymore. Moreover, we propose to learn a relationship-aware representation by jointly considering instances, attributes and relationships. By applying the representation-aware feature learned on VrR-VG, the performances of image captioning and visual question answering are systematically improved, which demonstrates the effectiveness of both our dataset and features embedding schema. Both our VrR-VG dataset and representation-aware features will be made publicly available soon.

Deep image embedding aims at learning a convolutional neural network (CNN) based mapping function that maps an image to a feature vector. The embedding quality is usually evaluated by the performance in image search tasks. Since very few users bother to open the second page search results, top-k precision mostly dominates the user experience and thus is one of the crucial evaluation metrics for the embedding quality. Despite being extensively studied, existing algorithms are usually based on heuristic observation without theoretical guarantee. Consequently, gradient descent direction on the training loss is mostly inconsistent with the direction of optimizing the concerned evaluation metric. This inconsistency certainly misleads the training direction and degrades the performance. In contrast to existing works, in this paper, we propose a novel deep image embedding algorithm with end-to-end optimization to top-k precision, the evaluation metric that is closely related to user experience. Specially, our loss function is constructed with wisely selected ``misplaced" images along the top k nearest neighbor decision boundary, so that the gradient descent update directly promotes the concerned metric, top-k precision. Further more, our theoretical analysis on the upper bounding and consistency properties of the proposed loss supports that minimizing our proposed loss is equivalent to maximizing top-k precision. Experiments show that our proposed algorithm outperforms all compared state-of-the-art deep image embedding algorithms on three benchmark datasets.

We describe an end-to-end method for recovering 3D human body mesh from single images and monocular videos. Different from the existing methods try to obtain all the complex 3D pose, shape, and camera parameters from one coupling feature, we propose a skeleton-disentangling based framework, which divides this task into multi-level spatial and temporal granularity in a decoupling manner. In spatial, we propose an effective and pluggable “disentangling the skeleton from the details” (DSD) module. It reduces the complexity and decouples the skeleton, which lays a good foundation for temporal modeling. In temporal, the self-attention based temporal convolution network is proposed to efficiently exploit the short and long-term temporal cues. Furthermore, an unsupervised adversarial training strategy, temporal shuffles and order recovery, is designed to promote the learning of motion dynamics. The proposed method outperforms the state-of-the-art 3D human mesh recovery methods by 15.4% MPJPE and 23.8% PA-MPJPE on Human3.6M. State-of-the-art results are also achieved on the 3D pose in the wild (3DPW) dataset without any fine-tuning. Especially, ablation studies demonstrate that skeleton-disentangled representation is crucial for better temporal modeling and generalization.

Face recognition has achieved significant progress with the growing scale of collected datasets, which empowers us to train strong convolutional neural networks (CNNs). While a variety of CNN architectures and loss functions have been devised recently, we still have a limited understanding of how to train the CNN models with the label noise inherent in existing face recognition datasets. To address this issue, this paper develops a novel co-mining strategy to effectively train on the datasets with noisy labels. Specifically, we simultaneously use the loss values as the cue to detect noisy labels, exchange the high-confidence clean faces to alleviate the errors accumulated issue caused by the sample-selection bias, and re-weight the predicted clean faces to make them dominate the discriminative model training in a mini-batch fashion. Extensive experiments by training on three popular datasets (\textit{i.e.}, CASIA-WebFace, MS-Celeb-1M and VggFace2) and testing on several benchmarks, including LFW, AgeDB, CFP, CALFW, CPLFW, RFW, and MegaFace, have demonstrated the effectiveness of our new approach over the state-of-the-art alternatives.

It is always well believed that parsing an image into constituent visual patterns would be helpful for understanding and representing an image. Nevertheless, there has not been evidence in support of the idea on describing an image with a natural-language utterance. In this paper, we introduce a new design to model a hierarchy from instance level (segmentation), region level (detection) to the whole image to delve into a thorough image understanding for captioning. Specifically, we present a HIerarchy Parsing (HIP) architecture that novelly integrates hierarchical structure into image encoder. Technically, an image decomposes into a set of regions and some of the regions are resolved into finer ones. Each region then regresses to an instance, i.e., foreground of the region. Such process naturally builds a hierarchal tree. A tree-structured Long Short-Term Memory (Tree-LSTM) network is then employed to interpret the hierarchal structure and enhance all the instance-level, region-level and image-level features. Our HIP is appealing in view that it is pluggable to any neural captioning models. Extensive experiments on COCO image captioning dataset demonstrate the superiority of HIP. More remarkably, HIP plus a top-down attention-based LSTM decoder increases CIDEr-D performance from 120.1% to 127.2% on COCO Karpathy test split. When further endowing instance-level and region-level features from HIP with semantic relation learnt through Graph Convolutional Networks (GCN), CIDEr-D is boosted up to 130.6%.

Selfie and cartoon are two popular artistic forms that are widely presented in our daily life. Despite the great progress in image translation/stylization, few techniques focus specifically on selfie cartoonization, since cartoon images usually contain artistic abstraction (e.g., large smoothing areas) and exaggeration (e.g., large/delicate eyebrows). In this paper, we address this problem by proposing a selfie cartoonization Generative Adversarial Network (scGAN), which mainly uses an attentive adversarial network (AAN) to emphasize specific facial regions and ignore low-level details. More specifically, we first design a cycle-like architecture to enable training with unpaired data. Then we design three losses from different aspects. A total variation loss is used to highlight important edges and contents in cartoon portraits. An attentive cycle loss is added to lay more emphasis on delicate facial areas such as eyes. In addition, a perceptual loss is included to eliminate artifacts and improve robustness of our method. Experimental results show that our method is capable of generating different cartoon styles and outperforms a number of state-of-the-art methods.

Current regression based text detectors mainly use fixed anchors, where scales and positions can not be changed during network training. As scene texts tend to have large variation in orientations, aspect ratios and sizes, fixed anchors are insufficient to cover all varieties. This paper proposes a novel text detector with learnable anchors, named LATD. LATD contains two prediction branches. One aims to refine scales and locations of anchors according to the characteristics of scene texts. The other one receives refined anchors as defaults and regresses their offsets to text regions. These two branches are optimized jointly without sacrifices much speed. Meanwhile, we explore the class-imbalance issue between texts and backgrounds, and replace softmax loss with focal loss. Extensive experiments on both oriented and horizontal benchmarks demonstrate the effectiveness of LATD with new state-of-the-art performance. By visualizing qualitative results, as expected, LATD provides more accurate locations and lower rate of missed detections.

Discovering social relations in images can make machines better interpret the behavior of human beings. However, automatically recognizing social relations in images is a challenging task due to the significant gap between the domains of visual content and social relation. Existing studies separately process various features such as faces expressions, body appearance, and contextual objects, thus they cannot comprehensively capture the multi-granularity semantics, such as scenes, regional cues of persons, and interactions among persons and objects. To bridge the domain gap, we propose a Multi-Granularity Reasoning framework for social relation recognition from images. The global knowledge and mid-level details are learned from the whole scene and the regions of persons and objects, respectively. Most importantly, we explore the fine-granularity pose keypoints of persons to discover the interactions among persons and objects. Specifically, the pose-guided Person-Object Graph and Person-Pose Graph are proposed to model the actions from persons to object and the interactions between paired persons, respectively. Based on the graphs, social relation reasoning is performed by graph convolutional networks. Finally, the global features and reasoned knowledge are integrated as a comprehensive representation for social relation recognition. Extensive experiments on two public datasets show the effectiveness of the proposed framework.

Image paragraph generation is the task of producing a coherent story (usually a paragraph) that describes the visual content of an image. The problem nevertheless is not trivial especially when there are multiple descriptive and diverse gists to be considered for paragraph generation, which often happens in real images. A valid question is how to encapsulate such gists/topics that are worthy of mention from an image, and then describe the image from one topic to another but holistically with a coherent structure. In this paper, we present a new design --- Convolutional Auto-Encoding (CAE) that purely employs convolutional and deconvolutional auto-encoding framework for topic modeling on the region-level features of an image. Furthermore, we propose an architecture, namely CAE plus Long Short-Term Memory (dubbed as CAE-LSTM), that novelly integrates the learnt topics in support of paragraph generation. Technically, CAE-LSTM capitalizes on a two-level LSTM-based paragraph generation framework with attention mechanism. The paragraph-level LSTM captures the inter-sentence dependency in a paragraph, while sentence-level LSTM is to generate one sentence which is conditioned on each learnt topic. Extensive experiments are conducted on Stanford image paragraph dataset, and superior results are reported when comparing to state-of-the-art approaches. More remarkably, CAE-LSTM increases CIDEr performance from 20.93% to 25.15%.

Unsupervised image-to-image translation is the task of translating an image from one domain to another in the absence of any paired training examples and tends to be more applicable to practical applications. Nevertheless, the extension of such synthesis from image-to-image to video-to-video is not trivial especially when capturing spatio-temporal structures in videos. The difficulty originates from the aspect that not only the visual appearance in each frame but also motion between consecutive frames should be realistic and consistent across transformation. This motivates us to explore both appearance structure and temporal continuity in video synthesis. In this paper, we present a new Motion-guided Cycle GAN, dubbed as Mocycle-GAN, that novelly integrates motion estimation into unpaired video translator. Technically, Mocycle-GAN capitalizes on three types of constrains: adversarial constraint discriminating between synthetic and real frame, cycle consistency encouraging an inverse translation on both frame and motion, and motion translation validating the transfer of motion between consecutive frames. Extensive experiments are conducted on video-to-labels and labels-to-video translation, and superior results are reported when comparing to state-of-the-art methods. More remarkably, we qualitatively demonstrate our Mocycle-GAN for both flower-to-flower and ambient condition transfer.

We demonstrate a video-to-animation translator, which can transform real-world video into cartoon or ink-wash animation in real-time. When users upload a video or record what they are seeing with the phone, the video-to-animation translator renders the live streaming video with cartoon or ink-wash animation style while maintaining the original contents. We formulate this task as video-to-video translation problem in the absence of any paired training examples, since the manual labeling of such paired video-animation data is cost-expensive and even unrealistic in practice. Technically, an unified unpaired video-to-video translator is utilized to explore both appearance structure and temporal continuity in video synthesis. As such, not only the visual appearance in each frame but also motion between consecutive frames are ensured to be realistic and consistent for video translation. Based on these technologies, our demonstration can be conducted on any videos in the wild and supports live video-to-animation translation, which engages users with the animated artistic expression of their life.

It has been well recognized that modeling human-object or object-object relations would be helpful for detection task. Nevertheless, the problem is not trivial especially when exploring the interactions between human actor, object and scene (collectively as human-context) to boost video action detectors. The difficulty originates from the aspect that reliable relations in a video should depend on not only short-term human-context relation in the present clip but also the temporal dynamics distilled over a long-range span of the video. This motivates us to capture both short-term and long-term relations in a video. In this paper, we present a new Long Short-Term Relation Networks, dubbed as LSTR, that novelly aggregates and propagates relation to augment features for video action detection. Technically, Region Proposal Networks (RPN) is remoulded to first produce 3D bounding boxes, i.e., tubelets, in each video clip. LSTR then models short-term human-context interactions within each clip through spatio-temporal attention mechanism and reasons long-term temporal dynamics across video clips via Graph Convolutional Networks (GCN) in a cascaded manner. Extensive experiments are conducted on four benchmark datasets, and superior results are reported when comparing to state-of-the-art methods.

This paper focuses on fine-grained human parsing in images. This is a very challenging task due to the diverse person appearance, semantic ambiguity of different body parts and clothing, and extremely small parsing targets. Although existing approaches can achieve significant improvement by pyramid feature learning, multi-level supervision, and joint learning with pose estimation, human parsing is still far from being solved. Different from existing approaches, we propose a Braiding Network, named as BraidNet, to learn complementary semantics and details for fine-grained human parsing. The BraidNet contains a two-stream braid-like architecture. The first stream is a semantic abstracting net with a deep yet narrow structure which can learn semantic knowledge by a hierarchy of fully convolution layers to overcome the challenges of diverse person appearance. To capture low-level details of small targets, the detail-preserving net is designed to exploit a shallow yet wide network without down-sampling, which can retain sufficient local structures for small objects. Moreover, we design a group of braiding modules across the two sub-nets, by which complementary information can be exchanged during end-to-end training. Besides, in the end of BraidNet, a Pairwise Hard Region Embedding strategy is propose to eliminate the semantic ambiguity of different body parts and clothing. Extensive experiments show that the proposed BraidNet achieves better performance than the state-of-the-art methods for fine-grained human parsing.

Multi-person pose tracking aims to jointly estimate and track multi-person keypoints in the unconstrained videos. The most popular solution to this task follows the tracking-by-detection strategy that relies on human detection and data association. While human detection has been boosted by deep learning, existing works mainly exploit several separated stages with hand-crafted metrics to realize data association, leading to great uncertainty and feeble adaption in complex scenes. To handle these problems, we propose an end-to-end pose-guided ovonic insight network (POINet) for the data association in multi-person pose tracking, which jointly learns feature extraction, similarity estimation, and identity assignment. Specifically, we design a pose-guided representation network to integrate pose information into hierarchical convolutional features, generating a pose-aligned person representation for person, which helps handle partial occlusions. Moreover, we propose an ovonic insight network to adaptively encode the cross-frame identity transformation, which can cope with the tough tracking cases of person leaving and entering the scene. In general, the proposed POINet provides a new insight to realize multi-person pose tracking in an end-to-end fashion. Extensive experiments conducted on the PoseTrack benchmark demonstrate that our POINet outperforms the state-of-the-art methods.

Merchandise categories inherently form a semantic hierarchy with different levels of concept abstraction, especially for fine-grained categories. This hierarchy encodes rich correlations among various categories across different levels, which can effectively regularize the semantic space and thus make prediction less ambiguous. However, previous studies of fine-grained image retrieval primarily focus on semantic similarities or visual similarities. In real application, merely using visual similarity may not satisfy the need of consumers to search merchandise with real-life images, e.g., given a red coat as query image, we might get red suit in recall results only based on visual similarity, since they are visually similar; But the users actually want coat rather than suit even the coat is with different color or texture attributes. We introduce this new problem based on photo shopping in real practice. That's why semantic information are integrated to regularize the margins to make "semantic" prior to "visual". To solve this new problem, we propose a hierarchical adaptive semantic-visual tree (ASVT) to depict the architecture of merchandise categories, which evaluates semantic similarities between different semantic levels and visual similarities within the same semantic class simultaneously. The semantic information satisfies the demand of consumers for similar merchandise with the query while the visual information optimize the correlations within the semantic class. At each level, we set different margins based on the semantic hierarchy and incorporate them as prior information to learn a fine-grained feature embedding. To evaluate our framework, we propose a new dataset named JDProduct, with hierarchical labels collected from actual image queries and official merchandise images on online shopping application. Extensive experimental results on the public CARS196 and CUB-200-2011 datasets demonstrate the superiority of our ASVT framework against compared state-of-the-art methods.

It is always well believed that Binary Neural Networks (BNNs) could drastically accelerate the inference efficiency by replacing the arithmetic operations in float-valued Deep Neural Networks (DNNs) with bit-wise operations. Nevertheless, there has not been open-source implementation in support of this idea on low-end ARM devices (e.g., mobile phones and embedded devices). In this work, we propose daBNN --- a super fast inference framework that implements BNNs on ARM devices. Several speed-up and memory refinement strategies for bit-packing, binarized convolution, and memory layout are uniquely devised to enhance inference efficiency. Compared to the recent open-source BNN inference framework, BMXNet, our daBNN is 7x~23x faster on a single binary convolution, and about 6x faster on Bi-Real Net 18 (a BNN variant of ResNet-18). The daBNN is a BSD-licensed inference framework, and its source code, sample projects and pre-trained models are available on-line: https://github.com/JDAI-CV/dabnn.

With advancement of various information processing and storage techniques, the scale of digital music collections has been growing at very fast speed during recent decades. To support high-quality content-based retrieval over such a large volume of music data, how to develop indexing structure with good effectiveness, efficiency and scalability becomes an important research issue. However, existing techniques mainly focus on improving query efficiency. Very few approaches have been proposed to address issues related to scalability and accuracy. In this study, we address the problem via introducing a novel indexing technique called effective music indexing framework (EMIF) to facilitate scalable and accurate music retrieval. It is designed based on a “classification-and-indexing” principle and consists of two main functionality modules: (1) music classification—a novel semantic-sensitive classification to identify an input song’s category and (2) indexing module—multiple local indexing structures, one for each semantic category to reduce query response time significantly. In particular, the classification model combining linear discriminative mixture model (LDMM) and advanced score fusion scheme has been applied to estimate category of music accurately. Layered architecture enables EMIF to enjoy superior scalability and efficiency. To evaluate the approach, a set of experimental studies has been carried out using two large music test collections and the results demonstrate various advantages of EMIF over state-of-the-art approaches including efficiency, scalability and effectiveness.

In this work, we investigate the problem of learning knowledge from the massive community-contributed images with rich weakly-supervised context information, which can benefit multiple image understanding tasks simultaneously, such as social image tag refinement and assignment, content-based image retrieval, tag-based image retrieval and tag expansion. Towards this end, we propose a Deep Collaborative Embedding (DCE) model to uncover a unified latent space for images and tags. The proposed method incorporates the end-to-end learning and collaborative factor analysis in one unified framework for the optimal compatibility of representation learning and latent space discovery. A nonnegative and discrete refined tagging matrix is learned to guide the end-to-end learning. To collaboratively explore the rich context information of social images, the proposed method integrates the weakly-supervised image-tag correlation, image correlation and tag correlation simultaneously and seamlessly. The proposed model is also extended to embed new tags in the uncovered space. To verify the effectiveness of the proposed method, extensive experiments are conducted on two widely-used social image benchmarks for multiple social image understanding tasks. The encouraging performance of the proposed method over the state-of-the-art approaches demonstrates its superiority.

This paper studies the subspace clustering problem. Given some data points approximately drawn from a union of subspaces, the goal is to group these data points into their underlying subspaces. Many subspace clustering methods have been proposed and among which sparse subspace clustering and low-rank representation are two representative ones. Despite the different motivations, we observe that many existing methods own the common block diagonal property, which possibly leads to correct clustering, yet with their proofs given case by case. In this work, we consider a general formulation and provide a unified theoretical guarantee of the block diagonal property. The block diagonal property of many existing methods falls into our special case. Second, we observe that many existing methods approximate the block diagonal representation matrix by using different structure priors, e.g., sparsity and low-rankness, which are indirect. We propose the first block diagonal matrix induced regularizer for directly pursuing the block diagonal matrix. With this regularizer, we solve the subspace clustering problem by Block Diagonal Representation (BDR), which uses the block diagonal structure prior. The BDR model is nonconvex and we propose an alternating minimization solver and prove its convergence. Experiments on real datasets demonstrate the effectiveness of BDR.

With the rapid growth of video content, video summarization, which focuses on automatically selecting important and informative parts from videos, is becoming increasingly crucial. However, the problem is challenging due to its subjectiveness. Previous research, which predominantly relies on manually designed criteria or resourcefully expensive human annotations, often fails to achieve satisfying results. We observe that the side information associated with a video (e.g., surrounding text such as titles, queries, descriptions, comments, and so on) represents a kind of human-curated semantics of video content. This side information, although valuable for video summarization, is overlooked in existing approaches. In this paper, we present a novel deep side semantic embedding (DSSE) model to generate video summaries by leveraging the freely available side information. The DSSE constructs a latent subspace by correlating the hidden layers of the two uni-modal autoencoders, which embed the video frames and side information, respectively. Specifically, by interactively minimizing the semantic relevance loss and the feature reconstruction loss of the two uni-modal autoencoders, the comparable common information between video frames and side information can be more completely learned. Therefore, their semantic relevance can be more effectively measured. Finally, semantically meaningful segments are selected from videos by minimizing their distances to the side information in the constructed latent subspace. We conduct experiments on two datasets (Thumb1K and TVSum50) and demonstrate the superior performance of DSSE to the several state-of-the-art approaches to video summarization.

Recognizing actions in videos is not a trivial task because video is an information-intensive media and includes multiple modalities. Moreover, on each modality, an action may only appear at some spatial regions, or only part of the temporal video segments may contain the action. A valid question is how to locate the attended spatial areas and selective video segments for action recognition. In this paper, we devise a general attention neural cell, called AttCell, that estimates the attention probability not only at each spatial location but also for each video segment in a temporal sequence. With AttCell, a unified Spatio-Temporal Attention Networks (STAN) is proposed in the context of multiple modalities. Specifically, STAN extracts the feature map of one convolutional layer as the local descriptors on each modality and pools the extracted descriptors with the spatial attention measured by AttCell as a representation of each segment. Then, we concatenate the representation on each modality to seek a consensus on the temporal attention, a priori, to holistically fuse the combined representation of video segments to the video representation for recognition. Our model differs from conventional deep networks, which focus on the attention mechanism, because our temporal attention provides a principled and global guidance across different modalities and video segments. Extensive experiments are conducted on four public datasets; UCF101, CCV, THUMOS14, and Sports-1M; our STAN consistently achieves superior results over several state-of-the-art techniques. More remarkably, we validate and demonstrate the effectiveness of our proposal when capitalizing on the different number of modalities.

In recent years, we have witnessed a flourish of social review websites. Internet users can easily share their experiences on some products and services with their friends. Therefore, measuring interpersonal influence becomes a popular method for recommender systems. However, traditional works are all based on external tangible activities, such as following, retweeting, mentioning, etc. In this paper, we explore user internal factors to measure his/her influence on a specific domain, namely, the social network on local businesses. The proposed user internal factors include user sentimental deviations and the review's reliability. The internal factors are not from explicit behavior but could help us to understand users. In addition, we utilize an attention mechanism that could auto-learn the weights of different factors. Through a case study on the Yelp dataset, we found that the proposed user internal factors on influence, that is, the proposed user sentimental deviations and the review's reliability, are effective in improving the accuracy of rating predictions.

One fundamental problem in image search is to learn the ranking functions (i.e., the similarity between query and image). Recent progress on this topic has evolved through two paradigms: the text-based model and image ranker learning. The former relies on image surrounding texts, making the similarity sensitive to the quality of textual descriptions. The latter may suffer from the robustness problem when human-labeled query-image pairs cannot represent user search intent precisely. We demonstrate in this article that the preceding two limitations can be well mitigated by learning a cross-view embedding that leverages click data. Specifically, a novel click-based Deep Structure-Preserving Embeddings with visual Attention (DSPEA) model is presented, which consists of two components: deep convolutional neural networks followed by image embedding layers for learning visual embedding, and a deep neural networks for generating query semantic embedding. Meanwhile, visual attention is incorporated at the top of the convolutional neural network to reflect the relevant regions of the image to the query. Furthermore, considering the high dimension of the query space, a new click-based representation on a query set is proposed for alleviating this sparsity problem. The whole network is end-to-end trained by optimizing a large margin objective that combines cross-view ranking constraints with in-view neighborhood structure preservation constraints. On a large-scale click-based image dataset with 11.7 million queries and 1 million images, our model is shown to be powerful for keyword-based image search with superior performance over several state-of-the-art methods and achieves, to date, the best reported NDCG@25 of 52.21%.

Despite the promising progress made in visual captioning and paragraphing, visual storytelling is still largely unexplored. This task is more challenging due to the difficulty in modeling an ordered photo sequence and in generating a relevant paragraph with expressive language style for storytelling. To deal with these challenges, we propose an Attribute-based Hierarchical Generative model with Reinforcement Learning and adversarial training (AHGRL). First, to model the ordered photo sequence and the complex story structure, we propose an attribute-based hierarchical generator. The generator incorporates semantic attributes to create more accurate and relevant descriptions. The hierarchical framework enables the generator to learn from the complex paragraph structure. Second, to generate story-style paragraphs, we design a language-style discriminator, which provides word-level rewards to optimize the generator by policy gradient. Third, we further consider the story generator and the reward critic as adversaries. The generator aims to create indistinguishable paragraphs to human-level stories, whereas the critic aims at distinguishing them and further improving the generator. Extensive experiments on the widely used dataset well demonstrate the advantages of the proposed method over state-of-the-art methods.

In recent years, Visual Question Answering (VQA) has attracted increasing attention due to its requirement on cross-modal understanding and reasoning of vision and language. VQA is proposed to automatically answer natural language questions with reference to a given image. VQA is challenging, because the reasoning process on a visual domain needs a full understanding of the spatial relationship, semantic concepts, as well as the common sense for a real image. However, most existing approaches jointly embed the abstract low-level visual features and high-level question features to infer answers. These works have limited reasoning ability due to the lack of modeling of the rich spatial context of regions, high-level semantics of images, and knowledge across multiple sources. To solve the challenges, we propose multi-source multi-level attention networks for visual question answering that can benefit both spatial inferences by visual attention on context-aware region representation and reasoning by semantic attention on concepts as well as external knowledge. Indeed, we learn to reason on image representation by question-guided attention at different levels across multiple sources, including region and concept level representation from image source as well as sentence level representation from the external knowledge base. First, we encode region-based middle-level outputs from Convolutional Neural Networks (CNNs) into spatially embedded representation by a multi-directional two-dimensional recurrent neural network and, further, locate the answer-related regions by Multiple Layer Perceptron as visual attention. Second, we generate semantic concepts from high-level semantics in CNNs and select those question-related concepts as concept attention. Third, we query semantic knowledge from the general knowledge base by concepts and selected question-related knowledge as knowledge attention. Finally, we jointly optimize visual attention, concept attention, knowledge attention, and question embedding by a softmax classifier to infer the final answer. Extensive experiments show the proposed approach achieved significant improvement on two very challenging VQA datasets.

Impressive image captioning results (i.e., an objective description for an image) are achieved with plenty of training pairs. In this paper, we take one step further to investigate the creation of narrative paragraph for a photo stream. This task is even more challenging due to the difficulty in modeling an ordered photo sequence and in generating a relevant paragraph with expressive language style for storytelling. The difficulty can even be exacerbated by the limited training data, so that existing approaches almost focus on search-based solutions. To deal with these challenges, we propose a sequence-to-sequence modeling approach with reinforcement learning and adversarial training. First, to model the ordered photo stream, we propose a hierarchical recurrent neural network as story generator, which is optimized by reinforcement learning with rewards. Second, to generate relevant and story-style paragraphs, we design the rewards with two critic networks, including a multi-modal and a language-style discriminator. Third, we further consider the story generator and reward critics as adversaries. The generator aims to create indistinguishable paragraphs to human-level stories, whereas the critics aim at distinguishing them and further improving the generator by policy gradient. Experiments on three widely-used datasets show the effectiveness, against state-of-the-art methods with relative increase of 20.2% by METEOR. We also show the subjective preference for the proposed approach over the baselines through a user study with 30 human subjects.

In this paper, we introduce the new ideas of augmenting Convolutional Neural Networks (CNNs) with Memory and learning to learn the network parameters for the unlabelled images on the fly in one-shot learning. Specifically, we present Memory Matching Networks (MM-Net) - a novel deep architecture that explores the training procedure, following the philosophy that training and test conditions must match. Technically, MM-Net writes the features of a set of labelled images (support set) into memory and reads from memory when performing inference to holistically leverage the knowledge in the set. Meanwhile, a Contextual Learner employs the memory slots in a sequential manner to predict the parameters of CNNs for unlabelled images. The whole architecture is trained by once showing only a few examples per class and switching the learning from minibatch to minibatch, which is tailored for one-shot learning when presented with a few examples of new categories at test time. Unlike the conventional one-shot learning approaches, our MM-Net could output one unified model irrespective of the number of shots and categories. Extensive experiments are conducted on two public datasets, i.e., Omniglot and miniImageNet, and superior results are reported when compared to state-of-the-art approaches. More remarkably, our MM-Net improves one-shot accuracy on Omniglot from 98.95% to 99.28% and from 49.21% to 53.37% on miniImageNet.

Automatically describing a video with natural language is regarded as a fundamental challenge in computer vision. The problem nevertheless is not trivial especially when a video contains multiple events to be worthy of mention, which often happens in real videos. A valid question is how to temporally localize and then describe events, which is known as "dense video captioning." In this paper, we present a novel framework for dense video captioning that unifies the localization of temporal event proposals and sentence generation of each proposal, by jointly training them in an end-to-end manner. To combine these two worlds, we integrate a new design, namely descriptiveness regression, into a single shot detection structure to infer the descriptive complexity of each detected proposal via sentence generation. This in turn adjusts the temporal locations of each event proposal. Our model differs from existing dense video captioning methods since we propose a joint and global optimization of detection and captioning, and the framework uniquely capitalizes on an attribute-augmented video captioning architecture. Extensive experiments are conducted on ActivityNet Captions dataset and our framework shows clear improvements when compared to the state-of-the-art techniques. More remarkably, we obtain a new record: METEOR of 12.96% on ActivityNet Captions official test set.

Unsupervised image translation, which aims in translating two independent sets of images, is challenging in discovering the correct correspondences without paired data. Existing works build upon Generative Adversarial Networks (GANs) such that the distribution of the translated images are indistinguishable from the distribution of the target set. However, such set-level constraints cannot learn the instance-level correspondences (e.g. aligned semantic parts in object transfiguration task). This limitation often results in false positives (e.g. geometric or semantic artifacts), and further leads to mode collapse problem. To address the above issues, we propose a novel framework for instance-level image translation by Deep Attention GAN (DA-GAN). Such a design enables DA-GAN to decompose the task of translating samples from two sets into translating instances in a highly-structured latent space. Specifically, we jointly learn a deep attention encoder, and the instance-level correspondences could be consequently discovered through attending on the learned instances. Therefore, the constraints could be exploited on both set-level and instance-level. Comparisons against several state-of-the-arts demonstrate the superiority of our approach, and the broad application capability, e.g, pose morphing, data augmentation, etc., pushes the margin of domain translation problem.1

The recent advances in deep neural networks have convincingly demonstrated high capability in learning vision models on large datasets. Nevertheless, collecting expert labeled datasets especially with pixel-level annotations is an extremely expensive process. An appealing alternative is to render synthetic data (e.g., computer games) and generate ground truth automatically. However, simply applying the models learnt on synthetic images may lead to high generalization error on real images due to domain shift. In this paper, we facilitate this issue from the perspectives of both visual appearance-level and representation-level domain adaptation. The former adapts source-domain images to appear as if drawn from the "style" in the target domain and the latter attempts to learn domain-invariant representations. Specifically, we present Fully Convolutional Adaptation Networks (FCAN), a novel deep architecture for semantic segmentation which combines Appearance Adaptation Networks (AAN) and Representation Adaptation Networks (RAN). AAN learns a transformation from one domain to the other in the pixel space and RAN is optimized in an adversarial learning manner to maximally fool the domain discriminator with the learnt source and target representations. Extensive experiments are conducted on the transfer from GTA5 (game videos) to Cityscapes (urban street scenes) on semantic segmentation and our proposal achieves superior results when comparing to state-of-the-art unsupervised adaptation techniques. More remarkably, we obtain a new record: mIoU of 47.5% on BDDS (drive-cam videos) in an unsupervised setting.

In Visual Question Answering, most existing approaches adopt the pipeline of representing an image via pre-trained CNNs, and then using the uninterpretable CNN features in conjunction with the question to predict the answer.Although such end-to-end models might report promising performance, they rarely provide any insight, apart from the answer, into the VQA process.In this work, we propose to break up the end-to-end VQA into two steps: explaining and reasoning, in an attempt towards a more explainable VQA by shedding light on the intermediate results between these two steps.To that end, we first extract attributes and generate descriptions as explanations for an image.Next, a reasoning module utilizes these explanations in place of the image to infer an answer.The advantages of such a breakdown include: (1) the attributes and captions can reflect what the system extracts from the image, thus can provide some insights for the predicted answer; (2) these intermediate results can help identify the inabilities of the image understanding or the answer inference part when the predicted answer is wrong.We conduct extensive experiments on a popular VQA dataset and our system achieves comparable performance with the baselines, yet with added benefits of explanability and the inherent ability to further improve with higher quality explanations.

In the modern e-commerce, the behaviors of customers contain rich information, e.g., consumption habits, the dynamics of preferences. Recently, session-based recommendationsare becoming popular to explore the temporal characteristics of customers' interactive behaviors. However, existing works mainly exploit the short-term behaviors without fully taking the customers' long-term stable preferences and evolutions into account. In this paper, we propose a novel Behavior-Intensive Neural Network (BINN) for next-item recommendation by incorporating both users' historical stable preferences and present consumption motivations. Specifically, BINN contains two main components, i.e., Neural Item Embedding, and Discriminative Behaviors Learning. Firstly, a novel item embedding method based on user interactions is developed for obtaining an unified representation for each item. Then, with the embedded items and the interactive behaviors over item sequences, BINN discriminatively learns the historical preferences and present motivations of the target users. Thus, BINN could better perform recommendations of the next items for the target users. Finally, for evaluating the performances of BINN, we conduct extensive experiments on two real-world datasets, i.e., Tianchi and JD. The experimental results clearly demonstrate the effectiveness of BINN compared with several state-of-the-art methods.

The recent advances in deep neural networks have demonstrated high capability in a wide variety of scenarios. Nevertheless, fine-tuning deep models in a new domain still requires a significant amount of labeled data despite expensive labeling efforts. A valid question is how to leverage the source knowledge plus unlabeled or only sparsely labeled target data for learning a new model in target domain. The core problem is to bring the source and target distributions closer in the feature space. In the paper, we facilitate this issue in an adversarial learning framework, in which a domain discriminator is devised to handle domain shift. Particularly, we explore the learning in the context of hashing problem, which has been studied extensively due to its great efficiency in gigantic data. Specifically, a novel Deep Domain Adaptation Hashing with Adversarial learning (DeDAHA) architecture is presented, which mainly consists of three components: a deep convolutional neural networks (CNN) for learning basic image/frame representation followed by an adversary stream on one hand to optimize the domain discriminator, and on the other, to interact with each domain-specific hashing stream for encoding image representation to hash codes. The whole architecture is trained end-to-end by jointly optimizing two types of losses, i.e., triplet ranking loss to preserve the relative similarity ordering in the input triplets and adversarial loss to maximally fool the domain discriminator with the learnt source and target feature distributions. Extensive experiments are conducted on three domain transfer tasks, including cross-domain digits retrieval, image to image and image to video transfers, on several benchmarks. Our DeDAHA framework achieves superior results when compared to the state-of-the-art techniques.

The Internet has emerged as the most effective and a highly popular medium for advertising. Current contextual advertising platforms need publishers to manually change the original structure of their Web pages and predefine the position and style of embedded ads. Although publishers spend significant effort optimizing their Web page layout, a large number of Web pages contain noticeable blank regions. We present an innovative stylewise advertising platform for contextual advertising, called PageSense. The “style” of Web pages refers to the visual appearance of a Web page, such as color and layout. PageSense aims to associate style-consistent ads with Web pages. It provides two advertising options: 1) If publishers predefine ad positions within Web pages, PageSense will analyze the page style and select ads, which are consistent with the Web page layout, and 2) if publishers impose no constraints for ad placement, PageSense will automatically detect blank regions, select the most nonintrusive region for ad insertion, associate color-consistent ads with the Web pages, and deliver them to blank regions without breaking the original Web page style. Our experiments have verified the effectiveness of PageSense as a complement to existing contextual advertising.

Recent studies have shown that a noticeable percentage of Web search traffic is about social events. While traditional Web sites can only show human-edited events, in this paper, we present a novel system to automatically detect events from search log data and generate storyboards where the events are arranged chronologically. We chose image search log as the resource for event mining, as search logs can directly reflect people's interests. To discover events from log data, we present a smooth nonnegative matrix factorization framework, which combines the information of query semantics, temporal correlations, search logs, and time continuity. Moreover, we consider the time factor to be an important element, since different events will develop in different time tendencies. In addition, to provide a media-rich and visually appealing storyboard, each event is associated with a set of representative photos arranged along a timeline. These relevant photos are automatically selected from image search results by analyzing image content features. We use celebrities as our test domain, which takes a large percentage of image search traffic. Experiments consisting of Web search traffic on 200 celebrities, for a period of six months, show very encouraging results compared with handcrafted editorial storyboards.

People are taking more photos than ever before in recent years. To effectively organize these personal photos, the photos are usually assigned to albums according to their events. An efficient way to manage our photos would be if we could recognize the events of the albums automatically. In this paper, we study the problem of recognizing events in personal photo albums. Recognizing events in photo albums is a new challenge since the contents of photos in albums are more complicated than in traditional single-photo tasks, since not all photos in an album are relevant to the event and a single photo in an album often fails to convey the meaningful event semantic behind the album. To solve this problem, we introduce an attention network to learn the representations of photo albums. Then, we adopt a hierarchical model to recognize events from coarse to fine using multigranular features. We evaluate our model on two real-world datasets consisting of personal albums; we find that our model achieves promising results.

Highlight detection from videos has been widely studied due to the fast growth of video contents. However, most existing approaches to highlight detection, either handcraft feature based or deep learning based, heavily rely on human-curated training data, which is very expensive to obtain and, thus, hinders the scalability to large datasets and unlabeled video categories. We observe that the largely available Web images can be applied as a weak supervision for highlight detection. For example, the top-ranked images in reference to the query “skiing” returned by a search engine may contain considerable positive samples of “skiing” highlights. Motivated by this observation, we propose a novel triplet deep ranking approach to video highlight detection using Web images as a weak supervision. The approach handles the relative preference of highlight scores between highlighting frames, nonhighlighting frames, and Web images by the triplet ranking constraints. Our approach can iteratively train two interdependent deep models (i.e., a triplet highlight model and a pairwise noise model) to deal with the noisy Web images in a single framework. We train the two models with relative preferences to generalize the capability regardless of the categories of training data. Therefore, our approach is fully category independent and exploits weakly supervised Web images. We evaluate our approach on two challenging datasets and achieve impressive results compared with the state-of-the-art pairwise ranking support vector machines, a robust recurrent autoencoder, and spatial deep convolution neural networks. We also empirically verify through cross-dataset evaluation that our category-independent model is fairly generalizable even if two different datasets do not share exactly the same categories.

Compared with person reidentification, which has attracted concentrated attention, vehicle reidentification is an important yet frontier problem in video surveillance and has been neglected by the multimedia and vision communities. Since most existing approaches mainly consider the general vehicle appearance for reidentification while overlooking the distinct vehicle identifier, such as the license plate number, they attain suboptimal performance. In this paper, we propose PROVID, a PROgressive Vehicle re-IDentification framework based on deep neural networks. In particular, our framework not only utilizes the multimodality data in large-scale video surveillance, such as visual features, license plates, camera locations, and contextual information, but also considers vehicle reidentification in two progressive procedures: coarse-to-fine search in the feature domain, and near-to-distant search in the physical space. Furthermore, to evaluate our progressive search framework and facilitate related research, we construct the VeRi dataset, which is the most comprehensive dataset from real-world surveillance videos. It not only provides large numbers of vehicles with varied labels and sufficient cross-camera recurrences but also contains license plate numbers and contextual information. Extensive experiments on the VeRi dataset demonstrate both the accuracy and efficiency of our progressive vehicle reidentification framework.

Semantically labeling every pixel in a video is a very challenging task as video is an information-intensive media with complex spatio-temporal dependence. We present in this paper a novel deep convolutional network architecture, called deep spatio-temporal fully convolutional networks (DST-FCN), which leverages both spatial and temporal dependencies among pixels and voxels by training them in an end-to-end manner. Specifically, we introduce a two-stream network by learning the deep spatio-temporal dependence, in which a 2D FCN followed by the convolutional long short-term memory (ConvLSTM) is employed on the pixel level and a 3-D FCN is exploited on the voxel level. Our model differs from conventional FCN in that it not only extends FCN by adding ConvLSTM on the pixel level for exploring long-term dependence, but also proposes 3-D FCN to enable voxel level prediction. On two benchmarks of A2D and CamVid, our DST-FCN achieves superior results to state-of-the-art techniques. More remarkably, we obtain to-date the best reported results: 45.0% per-label accuracy on A2D and 68.8% mean IoU on CamVid.

Large-scale image datasets and deep convolutional neural networks (DCNNs) are the two primary driving forces for the rapid progress in generic object recognition tasks in recent years. While lots of network architectures have been continuously designed to pursue lower error rates, few efforts are devoted to enlarging existing datasets due to high labeling costs and unfair comparison issues. In this article, we aim to achieve lower error rates by augmenting existing datasets in an automatic manner. Our method leverages both the web and DCNN, where the web provides massive images with rich contextual information, and DCNN replaces humans to automatically label images under the guidance of web contextual information. Experiments show that our method can automatically scale up existing datasets significantly from billions of web pages with high accuracy. The performance on object recognition tasks and transfer learning tasks have been significantly improved by using the automatically augmented datasets, which demonstrates that more supervisory information has been automatically gathered from the web. Both the dataset and models trained on the dataset have been made publicly available.

Automatic generation of natural language description for individual images (a.k.a. image captioning) has attracted extensive research attention. In this paper, we take one step further to investigate the generation of a paragraph to describe a photo stream for the purpose of storytelling. This task is even more challenging than individual image description due to the difficulty in modeling the large visual variance in an ordered photo collection and in preserving the long-term language coherence among multiple sentences. To deal with these challenges, we formulate the task as a sequence-to-sequence learning problem and propose a novel joint learning model by leveraging the semantic coherence in a photo stream. Specifically, to reduce visual variance, we learn a semantic space by jointly embedding each photo with its corresponding contextual sentence, so that the semantically related photos and their correlations are discovered. Then, to preserve language coherence in the paragraph, we learn a novel Bidirectional Attention-based Recurrent Neural Network (BARNN) model, which can attend on the discovered semantic relation to produce a sentence sequence and maintain its consistence with the photo stream. We integrate the two-step learning components into one single optimization formulation and train the network in an end-to-end manner. Experiments on three widely-used datasets (NYC/Disney/SIND) show that the proposed approach outperforms state-of-the-art methods with large margins for both retrieval and paragraph generation tasks. We also show the subjective preference of the machine-generated stories by the proposed approach over the baselines through a user study with 40 human subjects.

Recognizing fine-grained categories (e.g., bird species) is difficult due to the challenges of discriminative region localization and fine-grained feature learning. Existing approaches predominantly solve these challenges independently, while neglecting the fact that region detection and fine-grained feature learning are mutually correlated and thus can reinforce each other. In this paper, we propose a novel recurrent attention convolutional neural network (RA-CNN) which recursively learns discriminative region attention and region-based feature representation at multiple scales in a mutual reinforced way. The learning at each scale consists of a classification sub-network and an attention proposal sub-network (APN). The APN starts from full images, and iteratively generates region attention from coarse to fine by taking previous prediction as a reference, while the finer scale network takes as input an amplified attended region from previous scale in a recurrent way. The proposed RA-CNN is optimized by an intra-scale classification loss and an inter-scale ranking loss, to mutually learn accurate region attention and fine-grained representation. RA-CNN does not need bounding box/part annotations and can be trained end-to-end. We conduct comprehensive experiments and show that RA-CNN achieves the best performance in three fine-grained tasks, with relative accuracy gains of 3.3%, 3.7%, 3.8%, on CUB Birds, Stanford Dogs and Stanford Cars, respectively.

Automatically generating natural language descriptions of videos plays a fundamental challenge for computer vision community. Most recent progress in this problem has been achieved through employing 2-D and/or 3-D Convolutional Neural Networks (CNNs) to encode video content and Recurrent Neural Networks (RNNs) to decode a sentence. In this paper, we present Long Short-Term Memory with Transferred Semantic Attributes (LSTM-TSA) - a novel deep architecture that incorporates the transferred semantic attributes learnt from images and videos into the CNN plus RNN framework, by training them in an end-to-end manner. The design of LSTM-TSA is highly inspired by the facts that 1) semantic attributes play a significant contribution to captioning, and 2) images and videos carry complementary semantics and thus can reinforce each other for captioning. To boost video captioning, we propose a novel transfer unit to model the mutually correlated attributes learnt from images and videos. Extensive experiments are conducted on three public datasets, i.e., MSVD, M-VAD and MPIIMD. Our proposed LSTM-TSA achieves to-date the best published performance in sentence generation on MSVD: 52.8% and 74.0% in terms of BLEU@4 and CIDEr-D. Superior results are also reported on M-VAD and MPII-MD when compared to state-of-the-art methods.

Image captioning often requires a large set of training image-sentence pairs. In practice, however, acquiring sufficient training pairs is always expensive, making the recent captioning models limited in their ability to describe objects outside of training corpora (i.e., novel objects). In this paper, we present Long Short-Term Memory with Copying Mechanism (LSTM-C) - a new architecture that incorporates copying into the Convolutional Neural Networks (CNN) plus Recurrent Neural Networks (RNN) image captioning framework, for describing novel objects in captions. Specifically, freely available object recognition datasets are leveraged to develop classifiers for novel objects. Our LSTM-C then nicely integrates the standard word-by-word sentence generation by a decoder RNN with copying mechanism which may instead select words from novel objects at proper places in the output sentence. Extensive experiments are conducted on both MSCOCO image captioning and ImageNet datasets, demonstrating the ability of our proposed LSTM-C architecture to describe novel objects. Furthermore, superior results are reported when compared to state-of-the-art deep models.

This paper addresses the problem of joint detection and recounting of abnormal events in videos. Recounting of abnormal events, i.e., explaining why they are judged to be abnormal, is an unexplored but critical task in video surveillance, because it helps human observers quickly judge if they are false alarms or not. To describe the events in the human-understandable form for event recounting, learning generic knowledge about visual concepts (e.g., object and action) is crucial. Although convolutional neural networks (CNNs) have achieved promising results in learning such concepts, it remains an open question as to how to effectively use CNNs for abnormal event detection, mainly due to the environment-dependent nature of the anomaly detection. In this paper, we tackle this problem by integrating a generic CNN model and environment-dependent anomaly detectors. Our approach first learns CNN with multiple visual tasks to exploit semantic information that is useful for detecting and recounting abnormal events. By appropriately plugging the model into anomaly detectors, we can detect and recount abnormal events while taking advantage of the discriminative power of CNNs. Our approach outperforms the state-of-the-art on Avenue and UCSD Ped2 benchmarks for abnormal event detection and also produces promising results of abnormal event recounting.

Convolutional Neural Networks (CNN) have been regarded as a powerful class of models for image recognition problems. Nevertheless, it is not trivial when utilizing a CNN for learning spatio-temporal video representation. A few studies have shown that performing 3D convolutions is a rewarding approach to capture both spatial and temporal dimensions in videos. However, the development of a very deep 3D CNN from scratch results in expensive computational cost and memory demand. A valid question is why not recycle off-the-shelf 2D networks for a 3D CNN. In this paper, we devise multiple variants of bottleneck building blocks in a residual learning framework by simulating 3 x 3 x 3 convolutions with 1 × 3 × 3 convolutional filters on spatial domain (equivalent to 2D CNN) plus 3 × 1 × 1 convolutions to construct temporal connections on adjacent feature maps in time. Furthermore, we propose a new architecture, named Pseudo-3D Residual Net (P3D ResNet), that exploits all the variants of blocks but composes each in different placement of ResNet, following the philosophy that enhancing structural diversity with going deep could improve the power of neural networks. Our P3D ResNet achieves clear improvements on Sports-1M video classification dataset against 3D CNN and frame-based 2D CNN by 5.3% and 1.8%, respectively. We further examine the generalization performance of video representation produced by our pre-trained P3D ResNet on five different benchmarks and three different tasks, demonstrating superior performances over several state-of-the-art techniques.

Automatically describing an image with a natural language has been an emerging challenge in both fields of computer vision and natural language processing. In this paper, we present Long Short-Term Memory with Attributes (LSTM-A) - a novel architecture that integrates attributes into the successful Convolutional Neural Networks (CNNs) plus Recurrent Neural Networks (RNNs) image captioning framework, by training them in an end-to-end manner. Particularly, the learning of attributes is strengthened by integrating inter-attribute correlations into Multiple Instance Learning (MIL). To incorporate attributes into captioning, we construct variants of architectures by feeding image representations and attributes into RNNs in different ways to explore the mutual but also fuzzy relationship between them. Extensive experiments are conducted on COCO image captioning dataset and our framework shows clear improvements when compared to state-of-the-art deep models. More remarkably, we obtain METEOR/CIDEr-D of 25.5%/100.2% on testing data of widely used and publicly available splits in [10] when extracting image representations by GoogleNet and achieve superior performance on COCO captioning Leaderboard.

Recognizing fine-grained categories (e.g., bird species) highly relies on discriminative part localization and part-based fine-grained feature learning. Existing approaches predominantly solve these challenges independently, while neglecting the fact that part localization (e.g., head of a bird) and fine-grained feature learning (e.g., head shape) are mutually correlated. In this paper, we propose a novel part learning approach by a multi-attention convolutional neural network (MA-CNN), where part generation and feature learning can reinforce each other. MA-CNN consists of convolution, channel grouping and part classification sub-networks. The channel grouping network takes as input feature channels from convolutional layers, and generates multiple parts by clustering, weighting and pooling from spatially-correlated channels. The part classification network further classifies an image by each individual part, through which more discriminative fine-grained features can be learned. Two losses are proposed to guide the multi-task learning of channel grouping and part classification, which encourages MA-CNN to generate more discriminative parts from feature channels and learn better fine-grained features from parts in a mutual reinforced way. MA-CNN does not need bounding box/part annotation and can be trained end-to-end. We incorporate the learned parts from MA-CNN with part-CNN for recognition, and show the best performances on three challenging published fine-grained datasets, e.g., CUB-Birds, FGVC-Aircraft and Stanford-Cars.

Prediction of popularity has profound impact for social media, since it offers opportunities to reveal individual preference and public attention from evolutionary social systems. Previous research, although achieves promising results, neglects one distinctive characteristic of social data, i.e., sequentiality. For example, the popularity of online content is generated over time with sequential post streams of social media. To investigate the sequential prediction of popularity, we propose a novel prediction framework called Deep Temporal Context Networks (DTCN) by incorporating both temporal context and temporal attention into account. Our DTCN contains three main components, from embedding, learning to predicting. With a joint embedding network, we obtain a unified deep representation of multi-modal user-post data in a common embedding space. Then, based on the embedded data sequence over time, temporal context learning attempts to recurrently learn two adaptive temporal contexts for sequential popularity. Finally, a novel temporal attention is designed to predict new popularity (the popularity of a new user-post pair) with temporal coherence across multiple time-scales. Experiments on our released image dataset with about 600K Flickr photos demonstrate that DTCN outperforms state-of-the-art deep prediction algorithms, with an average of 21.51% relative performance improvement in the popularity prediction (Spearman Ranking Correlation).

Analyzing videos is one of the fundamental problems of computer vision and multimedia content analysis for decades. The task is very challenging as video is an information-intensive media with large variations and complexities. Thanks to the recent development of deep learning techniques, researchers in both computer vision and multimedia communities are now able to boost the performance of video analysis significantly and initiate new research directions to analyze video content. This tutorial will present recent advances under the umbrella of video understanding, which start from a unified deep learning toolkit--Microsoft Cognitive Toolkit (CNTK) that supports popular model types such as convolutional nets and recurrent networks, to fundamental challenges of video representation learning and video classification, recognition, and finally to an emerging area of video and language.

We are creating multimedia contents everyday and everywhere. While automatic content generation has played a fundamental challenge to multimedia community for decades, recent advances of deep learning have made this problem feasible. For example, the Generative Adversarial Networks (GANs) is a rewarding approach to synthesize images. Nevertheless, it is not trivial when capitalizing on GANs to generate videos. The difficulty originates from the intrinsic structure where a video is a sequence of visually coherent and semantically dependent frames. This motivates us to explore semantic and temporal coherence in designing GANs to generate videos. In this paper, we present a novel Temporal GANs conditioning on Captions, namely TGANs-C, in which the input to the generator network is a concatenation of a latent noise vector and caption embedding, and then is transformed into a frame sequence with 3D spatio-temporal convolutions. Unlike the naive discriminator which only judges pairs as fake or real, our discriminator additionally notes whether the video matches the correct caption. In particular, the discriminator network consists of three discriminators: video discriminator classifying realistic videos from generated ones and optimizes video-caption matching, frame discriminator discriminating between real and fake frames and aligning frames with the conditioning caption, and motion discriminator emphasizing the philosophy that the adjacent frames in the generated videos should be smoothly connected as in real ones. We qualitatively demonstrate the capability of our TGANs-C to generate plausible videos conditioning on the given captions on two synthetic datasets (SBMG and TBMG) and one real-world dataset (MSVD). Moreover, quantitative experiments on MSVD are performed to validate our proposal via Generative Adversarial Metric and human study.

Automatic generation of video caption is a challenging task as video is an information-intensive media with complex variations. Most existing methods, either based on language templates or sequence learning, have treated video as a flat data sequence while ignoring intrinsic multimodality nature. Observing that different modalities (e.g., frame, motion, and audio streams), as well as the elements within each modality, contribute differently to the sentence generation, we present a novel deep framework to boost video captioning by learning Multimodal Attention Long-Short Term Memory networks (MA-LSTM). Our proposed MA-LSTM fully exploits both multimodal streams and temporal attention to selectively focus on specific elements during the sentence generation. Moreover, we design a novel child-sum fusion unit in the MA-LSTM to effectively combine different encoded modalities to the initial decoding states. Different from existing approaches that employ the same LSTM structure for different modalities, we train modality-specific LSTM to capture the intrinsic representations of individual modalities. The experiments on two benchmark datasets (MSVD and MSR-VTT) show that our MA-LSTM significantly outperforms the state-of-the-art methods with 52.3 [email protected] and 70.4 CIDER-D metrics on MSVD dataset, respectively.

The ubiquitous mobile devices have led to the unprecedented growing of personal photo collections on the phone. One significant pain point of today's mobile users is instantly finding specific photos of what they want. Existing applications (e.g., Google Photo and OneDrive) have predominantly focused on cloud-based solutions, while leaving the client-side challenges (e.g., query formulation, photo tagging and search, etc.) unsolved. This considerably hinders user experience on the phone. In this paper, we present an innovative personal photo search system on the phone, which enables instant and accurate photo search by visual query suggestion and joint text-image hashing. Specifically, the system is characterized by several distinctive properties: 1) visual query suggestion (VQS) to facilitate the formulation of queries in a joint text-image form, 2) light-weight convolutional and sequential deep neural networks to extract representations for both photos and queries, and 3) joint text-image hashing (with compact binary codes) to facilitate binary image search and VQS. It is worth noting that all the components run on the phone with client optimization by deep learning techniques. We have collected 270 photo albums taken by 30 mobile users (corresponding to 37,000 personal photos) and conducted a series of field studies. We show that our system significantly outperforms the existing client-based solutions by 10 x in terms of search efficiency, and 92.3% precision in terms of search accuracy, leading to a remarkably better user experience of photo discovery on the phone.

We demonstrate a video captioning bot, named Seeing Bot, which can generate a natural language description about what it is seeing in near real time. Specifically, given a live streaming video, Seeing Bot runs two pre-learned and complementary captioning modules in parallel - one for generating image-level caption for each sampled frame, and the other for generating video-level caption for each sampled video clip. In particular, both the image and video captioning modules are boosted by incorporating semantic attributes which can enrich the generated descriptions, leading to human-level caption generation. A visual-semantic embedding model is then exploited to rank and select the final caption from the two parallel modules by considering the semantic relevance between video content and the generated captions. The Seeing Bot finally converts the generated description to speech and sends the speech to an end user via an earphone. Our demonstration is conducted on any videos in the wild and supports live video captioning.

Hashing has been a widely-adopted technique for nearest neighbor search in large-scale image retrieval tasks. Recent research has shown that leveraging supervised information can lead to high quality hashing. However, the cost of annotating data is often an obstacle when applying supervised hashing to a new domain. Moreover, the results can suffer from the robustness problem as the data at training and test stage may come from different distributions. This paper studies the exploration of generating synthetic data through semi-supervised generative adversarial networks (GANs), which leverages largely unlabeled and limited labeled training data to produce highly compelling data with intrinsic invariance and global coherence, for better understanding statistical structures of natural data. We demonstrate that the above two limitations can be well mitigated by applying the synthetic data for hashing. Specifically, a novel deep semantic hashing with GANs (DSH-GANs) is presented, which mainly consists of four components: a deep convolution neural networks (CNN) for learning image representations, an adversary stream to distinguish synthetic images from real ones, a hash stream for encoding image representations to hash codes and a classification stream. The whole architecture is trained end-to-end by jointly optimizing three losses, i.e., adversarial loss to correct label of synthetic or real for each sample, triplet ranking loss to preserve the relative similarity ordering in the input real-synthetic triplets and classification loss to classify each sample accurately. Extensive experiments conducted on both CIFAR-10 and NUS-WIDE image benchmarks validate the capability of exploiting synthetic images for hashing. Our framework also achieves superior results when compared to state-of-the-art deep hash models.

Feature selection (FS) is an important technique in machine learning and data mining, especially for large-scale high-dimensional data. Most existing studies have been restricted to batch learning, which is often inefficient and poorly scalable when handling big data in real world. As real data may arrive sequentially and continuously, batch learning has to retrain the model for the new coming data, which is very computationally intensive. Online feature selection (OFS) is a promising new paradigm that is more efficient and scalable than batch learning algorithms. However, existing online algorithms usually fall short in their inferior efficacy. In this article, we present a novel second-order OFS algorithm that is simple yet effective, very fast and extremely scalable to deal with large-scale ultra-high dimensional sparse data streams. The basic idea is to exploit the second-order information to choose the subset of important features with high confidence weights. Unlike existing OFS methods that often suffer from extra high computational cost, we devise a novel algorithm with a MaxHeap-based approach, which is not only more effective than the existing first-order algorithms, but also significantly more efficient and scalable. Our extensive experiments validated that the proposed technique achieves highly competitive accuracy as compared with state-of-the-art batch FS methods, meanwhile it consumes significantly less computational cost that is orders of magnitude lower. Impressively, on a billion-scale synthetic dataset (1-billion dimensions, 1-billion non-zero features, and 1-million samples), the proposed algorithm takes less than 3 minutes to run on a single PC.

The Internet has extended the physical boundary of people's social circles to manage an inordinate number of online friends. It is recognized that only a fraction of these online friends are also known with each other in offline circumstances, i.e., the offline friends. An important type of offline friend, onsite offline friend, is defined and addressed in this paper. We explores the possibility of utilizing users' online photo sharing-related behaviors and network topologies to analyze and distinguish between online and onsite offline friendships. Different from traditional social science studies which rely on survey-based data, we employ users' tagged people on the shared Instagram photos as the ground-truth for onsite offline friends. This enables a large-scale and objective analysis and experimental evaluation, which compares between different factors and identifies the features that are key to onsite offline friend identification.

Time information plays a crucial role on social media popularity. Existing research on popularity prediction, effective though, ignores temporal information which is highly related to user-item associations and thus often results in limited success. An essential way is to consider all these factors (user, item, and time), which capture the dynamic nature of photo popularity. In this paper, we present a novel approach to factorize the popularity into user-item context and time-sensitive context for exploring the mechanism of dynamic popularity. The user-item context provides a holistic view of popularity, while the time-sensitive context captures the temporal dynamics nature of popularity. Accordingly, we develop two kinds of time-sensitive features, including user activeness variability and photo prevalence variability. To predict photo popularity, we propose a novel framework named Multi-scale Temporal Decomposition (MTD), which decomposes the popularity matrix in latent spaces based on contextual associations. Specifically, the proposed MTD models time-sensitive context on different time scales, which is beneficial to automatically learn temporal patterns. Based on the experiments conducted on a real-world dataset with 1.29M photos from Flickr, our proposed MTD can achieve the prediction accuracy of 79.8% and outperform the best three state-of-the-art methods with a relative improvement of 9.6% on average.

Video concept learning often requires a large set oftraining samples. In practice, however, acquiring noise-free training labels with sufficient positive examples is very expensive. A plausible solution for training data collection is by sampling from the vast quantities of images and videos on the Web. Such a solution is motivated by the assumption that the retrieved images or videos are highly correlated with the query. Still, a number ofchallenges remain. First, Web videos are often untrimmed. Thus, only parts of the videos are relevant to the query. Second, the retrieved Web images are always highly relevant to the issued query. However, thoughtlessly utilizing the images in the video domain may even hurt the performance due to the well-known semantic drift and domain gap problems. As a result, a valid question is how Web images and videos interact for video concept learning. In this paper, we propose a Lead-Exceed Neural Network (LENN), which reinforces the training on Web images and videos in a curriculum manner. Specifically, the training proceeds by inputting frames of Web videos to obtain a network. The Web images are then filtered by the learnt network and the selected images are additionally fed into the network to enhance the architecture and further trim the videos. In addition, Long Short-Term Memory (LSTM) can be applied on the trimmed videos to explore temporal information. Encouraging results are reported on UCFIOl, TRECVID 2013 and 2014 MEDTest in the context ofboth action recognition and event detection. Without using human annotated exemplars, our proposed LENN can achieve 74.4% accuracy on UCFIOI dataset.

Automatically describing video content with natural language is a fundamental challenge of computer vision. Re-current Neural Networks (RNNs), which models sequence dynamics, has attracted increasing attention on visual interpretation. However, most existing approaches generate a word locally with the given previous words and the visual content, while the relationship between sentence semantics and visual content is not holistically exploited. As a result, the generated sentences may be contextually correct but the semantics (e.g., subjects, verbs or objects) are not true. This paper presents a novel unified framework, named Long Short-Term Memory with visual-semantic Embedding (LSTM-E), which can simultaneously explore the learning of LSTM and visual-semantic embedding. The former aims to locally maximize the probability of generating the next word given previous words and visual content, while the latter is to create a visual-semantic embedding space for enforcing the relationship between the semantics of the entire sentence and visual content. The experiments on YouTube2Text dataset show that our proposed LSTM-E achieves to-date the best published performance in generating natural sentences: 45.3% and 31.0% in terms of BLEU@4 and METEOR, respectively. Superior performances are also reported on two movie description datasets (M-VAD and MPII-MD). In addition, we demonstrate that LSTM-E outperforms several state-of-the-art techniques in predicting Subject-Verb-Object (SVO) triplets.

While there has been increasing interest in the task of describing video with natural language, current computer vision algorithms are still severely limited in terms of the variability and complexity of the videos and their associated language that they can recognize. This is in part due to the simplicity of current benchmarks, which mostly focus on specific fine-grained domains with limited videos and simple descriptions. While researchers have provided several benchmark datasets for image captioning, we are not aware of any large-scale video description dataset with comprehensive categories yet diverse video content. In this paper we present MSR-VTT (standing for "MSRVideo to Text") which is a new large-scale video benchmark for video understanding, especially the emerging task of translating video to text. This is achieved by collecting 257 popular queries from a commercial video search engine, with 118 videos for each query. In its current version, MSR-VTT provides 10K web video clips with 41.2 hours and 200K clip-sentence pairs in total, covering the most comprehensive categories and diverse visual content, and representing the largest dataset in terms of sentence and vocabulary. Each clip is annotated with about 20 natural sentences by 1,327 AMT workers. We present a detailed analysis of MSR-VTT in comparison to a complete set of existing datasets, together with a summarization of different state-of-the-art video-to-text approaches. We also provide an extensive evaluation of these approaches on this dataset, showing that the hybrid Recurrent Neural Networkbased approach, which combines single-frame and motion representations with soft-attention pooling strategy, yields the best generalization capability on MSR-VTT.

The emergence of wearable devices such as portable cameras and smart glasses makes it possible to record life logging first-person videos. Browsing such long unstructured videos is time-consuming and tedious. This paper studies the discovery of moments of user's major or special interest (i.e., highlights) in a video, for generating the summarization of first-person videos. Specifically, we propose a novel pairwise deep ranking model that employs deep learning techniques to learn the relationship between high-light and non-highlight video segments. A two-stream network structure by representing video segments from complementary information on appearance of video frames and temporal dynamics across frames is developed for video highlight detection. Given a long personal video, equipped with the highlight detection model, a highlight score is assigned to each segment. The obtained highlight segments are applied for summarization in two ways: video time-lapse and video skimming. The former plays the highlight (non-highlight) segments at low (high) speed rates, while the latter assembles the sequence of segments with the highest scores. On 100 hours of first-person videos for 15 unique sports categories, our highlight detection achieves the improvement over the state-of-the-art RankSVM method by 10.5% in terms of accuracy. Moreover, our approaches produce video summary with better quality by a user study from 35 human subjects.

Digital storytelling applications are playing an increasingly important role in people's daily life. In contemporary storytelling applications such as PowerPoint presentation and macro/micro blogs, good presentation images are always highly desired by content creators to boost their presentation in an intuitive and attractive way. Existing studies, however, have not yet addressed the challenging problem of how to select the most appropriate presentation images for storytelling. In this paper, we formulate this problem of presentation image suggestion (given a textual query) as selecting images by maximizing visual and semantic diversity from web image search results of suggested queries. The proposed framework consists of two novel components: 1) click-through-based query suggestion, which is designed to suggest textual queries by searching relevant queries in a constructed query graph that can reflect diverse aspects of a given query, and 2) query-based image selection, which selects the most appropriate presentation images by keeping semantic relevance while maximizing visual diversity and quality, using a novel model based on Conditional Random Field (CRF) by individual and correlation characters. We evaluate the proposed approach by comparing with several baselines and a thorough subjective survey. The evaluations show inspiring results using the proposed approach for automatic suggestion of images for storytelling.

Recognizing actions in videos is a challenging task as video is an information-intensive media with complex variations. Most existing methods have treated video as a flat data sequence while ignoring the intrinsic hierarchical structure of the video content. In particular, an action may span different granularities in this hierarchy including, from small to large, a single frame, consecutive frames (motion), a short clip, and the entire video. In this paper, we present a novel framework to boost action recognition by learning a deep spatio-temporal video representation at hierarchical multi-granularity. Specifically, we model each granularity as a single stream by 2D (for frame and motion streams) or 3D (for clip and video streams) convolutional neural networks (CNNs). The framework therefore consists of multi-stream 2D or 3D CNNs to learn both the spatial and temporal representations. Furthermore, we employ the Long Short-Term Memory (LSTM) networks on the frame, motion, and clip streams to exploit long-term temporal dynamics. With a softmax layer on the top of each stream, the classification scores can be predicted from all the streams, followed by a novel fusion scheme based on the multi-granular score distribution. Our networks are learned in an end-to-end fashion. On two video action benchmarks of UCF101 and HMDB51, our framework achieves promising performance compared with the state-of-the-art.

We demonstrate a video chatbot, which can generate human-level emotional comments referring to the videos shared by users and trigger a conversation with users. Our video chatbot performs a large-scale similar video search to find visually similar videos w.r.t. a given video using approximate nearest-neighbor search. Then, the comments associated with the searched similar videos are ranked by learning a deep multi-view embedding space for modeling video content, visual sentiment and textual comments. The top ranked comments are selected as responses to the given video and trigger the succeeding text-based chat between users and the chatbot. The demonstration is conducted on a newly collected dataset with over 102K videos and 10.6M comments. Moreover, our video chatbot has great potential to increase live social interactions.

Video has become a predominant social media for the booming live interactions. Automatic generation of emotional comments to a video has great potential to significantly increase user engagement in many socio-video applications (e.g., chat bot). Nevertheless, the problem of video commenting has been overlooked by the research community. The major challenges are that the generated comments are to be not only as natural as those from human beings, but also relevant to the video content. We present in this paper a novel two-stage deep learning-based approach to automatic video commenting. Our approach consists of two components. The first component, similar video search, efficiently finds the visually similar videos w.r.t. a given video using approximate nearest-neighbor search based on the learned deep video representations, while the second dynamic ranking effectively ranks the comments associated with the searched similar videos by learning a deep multi-view embedding space. For modeling the emotional view of videos, we incorporate visual sentiment, video content, and text comments into the learning of the embedding space. On a newly collected dataset with over 102K videos and 10.6M comments, we demonstrate that our approach outperforms several state-of-the-art methods and achieves human-level video commenting.

Person re-identification aims at identifying a certain person across non-overlapping multi-camera networks. It is a fundamental and challenging task in automated video surveillance. Most existing researches mainly rely on hand-crafted features, resulting in unsatisfactory performance. In this paper, we propose a multi-scale triplet convolutional neural network which captures visual appearance of a person at various scales. We propose to optimize the network parameters by a comparative similarity loss on massive sample triplets, addressing the problem of small training set in person re-identification. In particular, we design a unified multi-scale network architecture consisting of both deep and shallow neural networks, towards learning robust and effective features for person re-identification under complex conditions. Extensive evaluation on the real-world Market-1501 dataset have demonstrated the effectiveness of the proposed approach.

The evolution of social media popularity exhibits rich temporality, i.e., popularities change over time at various levels of temporal granularity. This is influenced by temporal variations of public attentions or user activities. For example, popularity patterns of street snap on Flickr are observed to depict distinctive fashion styles at specific time scales, such as season-based periodic fluctuations for Trench Coat or one-off peak in days for Evening Dress. However, this fact is often overlooked by existing research of popularity modeling. We present the first study to incorporate multiple time-scale dynamics into predicting online popularity. We propose a novel computational framework in the paper, named Multi-scale Temporalization, for estimating popularity based on multi-scale decomposition and structural reconstruction in a tensor space of user, post, and time by joint low-rank constraints. By considering the noise caused by context inconsistency, we design a data rearrangement step based on context aggregation as preprocessing to enhance contextual relevance of neighboring data in the tensor space. As a result, our approach can leverage multiple levels of temporal characteristics and reduce the noise of data decomposition to improve modeling effectiveness. We evaluate our approach on two large-scale Flickr image datasets with over 1.8 million photos in total, for the task of popularity prediction. The results show that our approach significantly outperforms state-of-the-art popularity prediction techniques, with a relative improvement of 10.9%-47.5% in terms of prediction accuracy.

Big data increasingly benefit both research and industrial area such as health care, finance service and commercial recommendation. This paper presents a personalized travel sequence recommendation from both travelogues and community contributed photos and the heterogeneous metadata (e.g., tags, geo-location, and date taken) associated with these photos. Unlike most existing travel recommendation approaches, our approach is not only personalized to user's travel interest but also able to recommend a travel sequence rather than individual Points of Interest (POIs). Topical package space including representative tags, the distributions of cost, visiting time and visiting season of each topic, is mined to bridge the vocabulary gap between user travel preference and travel routes. We take advantage of the complementary of two kinds of social media: travelogue and community contributed photos. We map both user's and routes' textual descriptions to the topical package space to get user topical package model and route topical package model (i.e., topical interest, cost, time and season). To recommend personalized POI sequence, first, famous routes are ranked according to the similarity between user package and route package. Then top ranked routes are further optimized by social similar users' travel records. Representative images with viewpoint and seasonal diversity of POIs are shown to offer a more comprehensive impression. We evaluate our recommendation system on a collection of 7 million Flickr images uploaded by 7,387 users and 24,008 travelogues covering 864 travel POIs in nine famous cities, and show its effectiveness. We also contribute a new dataset with more than 200 K photos with heterogeneous metadata in nine famous cities.

An ideal thumbnail generator should effectively condense unimportant regions and keep the important content undeformed, completed, and at a proper scale, i.e., accuracy, completeness, and sufficiency. Each retargeting method has its own advantage for resizing arbitrary images. However, they often ignore the completeness and sufficiency for information presentation in thumbnails. In this paper, we formulate thumbnail generation as an image content condensation problem and propose a unified grid optimization framework to fuse multiple operators. From the view of accuracy, completeness, and sufficiency for information presentation, we exploit complementary relationships among three condensation operators and fuse them into a unified grid-based convex programming problem, which could be solved simultaneously and efficiently through numerical optimization. Besides warping energy to preserve the geometric structure of important objects, we put forward two grid-based energy terms to keep the completeness of important objects and retain them at a proper size. Finally, an adaptive procedure is proposed to dynamically adjust the contribution of loss functions for achieving optimal content condensation. Both qualitative and quantitative comparison results demonstrate that the proposed method achieves an excellent tradeoff among accuracy, completeness, and sufficiency of information preservation. The experimental results show that our approach is obviously superior to the state-of-the-art techniques.

The visualization of an image collection is the process of displaying a collection of images on a screen under some specific layout requirements. This paper focuses on an important problem that is not well addressed by the previous methods: visualizing image collections into arbitrary layout shapes while arranging images according to user-defined semantic or visual correlations (e.g., color or object category). To this end, we first propose a property-based tree construction scheme to organize images of a collection into a tree structure according to user-defined properties. In this way, images can be adaptively placed with the desired semantic or visual correlations in the final visualization layout. Then, we design a two-step visualization optimization scheme to further optimize image layouts. As a result, multiple layout effects including layout shape and image overlap ratio can be effectively controlled to guarantee a satisfactory visualization. Finally, we also propose a tree-transfer scheme such that visualization layouts can be adaptively changed when users select different "images of interest." We demonstrate the effectiveness of our proposed approach through the comparisons with state-of-the-art visualization techniques.

This paper addresses the problem of detecting coherent motions in crowd scenes and presents its two applications in crowd scene understanding: semantic region detection and recurrent activity mining. It processes input motion fields (e.g., optical flow fields) and produces a coherent motion field named thermal energy field. The thermal energy field is able to capture both motion correlation among particles and the motion trends of individual particles, which are helpful to discover coherency among them. We further introduce a two-step clustering process to construct stable semantic regions from the extracted time-varying coherent motions. These semantic regions can be used to recognize pre-defined activities in crowd scenes. Finally, we introduce a cluster-and-merge process, which automatically discovers recurrent activities in crowd scenes by clustering and merging the extracted coherent motions. Experiments on various videos demonstrate the effectiveness of our approach.

In image search re-ranking, besides the well-known semantic gap, intent gap, which is the gap between the representation of users' query/demand and the real intent of the users, is becoming a major problem restricting the development of image retrieval. To reduce human effects, in this paper, we use image click-through data, which can be viewed as the implicit feedback from users, to help overcome the intention gap, and further improve the image search performance. Generally, the hypothesis-visually similar images should be close in a ranking list-and the strategy-images with higher relevance should be ranked higher than others-are widely accepted. To obtain satisfying search results, thus, image similarity and the level of relevance typicality are determinate factors correspondingly. However, when measuring image similarity and typicality, conventional re-ranking approaches only consider visual information and initial ranks of images, while overlooking the influence of click-through data. This paper presents a novel re-ranking approach, named spectral clustering re-ranking with click-based similarity and typicality. First, to learn an appropriate similarity measurement, we propose click-based multi-feature similarity learning algorithm, which conducts metric learning based on click-based triplets selection, and integrates multiple features into a unified similarity space via multiple kernel learning. Then, based on the learnt click-based image similarity measure, we conduct spectral clustering to group visually and semantically similar images into same clusters, and get the final re-rank list by calculating click-based clusters typicality and within-clusters click-based image typicality in descending order. Our experiments conducted on two real-world query-image data sets with diverse representative queries show that our proposed re-ranking approach can significantly improve initial search results, and outperform several existing re-ranking approaches.

It is an important yet challenging task for investors to determine the most suitable type of shop (e.g., restaurant, fashion) for a newly opened store. Traditional ways are predominantly field surveys and empirical estimation, which are not effective as they lack shop-related data. As social media and location-based services (LBS) are becoming more and more pervasive, user-generated data from these platforms are providing rich information not only about individual consumption experiences, but also about shop attributes. In this paper, we investigate the recommendation of shop types for a given location, by leveraging heterogeneous data that are mainly historical user preferences and location context from social media and LBS. Our goal is to select the most suitable shop type, seeking to maximize the number of customers served from a candidate set of types. We propose a novel bias learning matrix factorization method with feature fusion for shop popularity prediction. Features are defined and extracted from two perspectives: location, where features are closely related to location characteristics, and commercial, where features are about the relationships between shops in the neighborhood. Experimental results show that the proposed method outperforms state-of-the-art solutions.

Peer-to-peer networking offers a scalable solution for sharing multimedia data across the network. With a large amount of visual data distributed among different nodes, it is an important but challenging issue to perform content-based retrieval in peer-to-peer networks. While most of the existing methods focus on indexing high dimensional visual features and have limitations of scalability, in this paper we propose a scalable approach for content-based image retrieval in peer-to-peer networks by employing the bag-of-visual-words model. Compared with centralized environments, the key challenge is to efficiently obtain a global codebook, as images are distributed across the whole peer-to-peer network. In addition, a peer-to-peer network often evolves dynamically, which makes a static codebook less effective for retrieval tasks. Therefore, we propose a dynamic codebook updating method by optimizing the mutual information between the resultant codebook and relevance information, and the workload balance among nodes that manage different codewords. In order to further improve retrieval performance and reduce network cost, indexing pruning techniques are developed. Our comprehensive experimental results indicate that the proposed approach is scalable in evolving and distributed peer-to-peer networks, while achieving improved retrieval accuracy.

Nowadays, with the boom of social media and e-commerce, more and more people prefer to share their consumption experiences and rate services on review sites. Much research has focused on personalized recommendation. However, quality of service also plays an important role in recommender systems, and it is the main concern of this paper. An overall rating that indicates the popular view usually represents the evaluation. There are some challenges when we do not have enough review information to extract public opinion. Take, for example, a movie for which one user rates a two star rating, and another rates a five star rating. In this case, it is difficult to conduct a quality evaluation fairly. However, it is possible to be improved with the help of big social users' contextual information. In this paper, we propose a model to conduct service quality evaluation by improving overall rating of services using an empirical methodology. We use the concept of user rating's confidence, which denotes the trustworthiness of user ratings. First, entropy is utilized to calculate user ratings' confidence. Second, we further explore spatial-temporal features and review sentimental features of user ratings to constrain their confidences. Last, we fuse them into a unified model to calculate an overall confidence, which is utilized to perform service quality evaluation. Extensive experiments implemented on Yelp and Douban Movie datasets demonstrate the effectiveness of our model.

With the ever-increasing use of smartphones and digital cameras, people are now able to take photos anywhere and anytime. Most of these photos simply end up stored in the cloud without further interaction. This occurs because we lack intelligent services to organize these personal photos well. Therefore, there is an urgent need for such a system to enable people to relive their memories by turning their photos into stories. This paper presents a storytelling system named Monet, which automatically creates interesting stories from personal photos by mimicking cinematic knowledge based on a set of predesigned editing styles. The system consists of two stages: photo summarization, which selects a subset of the “best” photos to represent a photo collection, and story remixing, which generates a stylish music video from the selected photos. During photo summarization, photos are grouped into events based on multimodal features (time and location). The “best” photos are then selected according to visual quality, event representativeness, and diversity. The second stage, story remixing, automatically selects an appropriate theme-dependent editing style based on the photo content. Each selected photo is converted to a video clip by applying a virtual camera with appropriate motions. A series of video effects, color filters, shapes, and transitions are then applied to the video clips according to cinematic rules. The generated video is finally multiplexed with a music clip to generate the story. Evaluations show that our system achieves superior performance to state-of-the-art photo event detection and story generation systems.

Visual-textual presentation layout (e.g., digital magazine cover, poster, Power Point slides, and any other rich media), which combines beautiful image and overlaid readable texts, can result in an eye candy touch to attract users’ attention. The designing of visual-textual presentation layout is therefore becoming ubiquitous in both commercially printed publications and online digital magazines. However, handcrafting aesthetically compelling layouts still remains challenging for many small businesses and amateur users. This article presents a system to automatically generate visual-textual presentation layouts by investigating a set of aesthetic design principles, through which an average user can easily create visually appealing layouts. The system is attributed with a set of topic-dependent layout templates and a computational framework integrating high-level aesthetic principles (in a top-down manner) and low-level image features (in a bottom-up manner). The layout templates, designed with prior knowledge from domain experts, define spatial layouts, semantic colors, harmonic color models, and font emotion and size constraints. We formulate the typography as an energy optimization problem by minimizing the cost of text intrusion, the utility of visual space, and the mismatch of information importance in perception and semantics, constrained by the automatically selected template and further preserving color harmonization. We demonstrate that our designs achieve the best reading experience compared with the reimplementation of parts of existing state-of-the-art designs through a series of user studies.

Advanced networking technologies and massive online social media have stimulated a booming growth of travel heterogeneous information in recent years. By employing such information, smart travel guide systems, such as landmark ranking systems, have been proposed to offer diverse online travel services. It is essential for a landmark ranking system to structure, analyze, and search the travel heterogeneous information to produce human-expected results. Therefore, currently the most fundamental yet challenging problems can be concluded: 1) how to fuse heterogeneous tourism information and 2) how to model landmark ranking. In this paper, a novel landmark search system is introduced based on a newly designed heterogeneous information fusion scheme and a query-dependent landmark ranking strategy. Different from the existing travel guide systems, the proposed system can effectively combine the heterogeneous information from multimodality media into a landmark reranking list via a user's query. Experimental results conducted on a large travel information collection illustrate the advantages of the proposed system in terms of both effectiveness and efficiency.

Most state-of-the-art image retrieval and recommendation systems predominantly focus on individual images. In contrast, socially curated image collections, condensing distinctive yet coherent images into one set, are largely overlooked by the research communities. In this paper, we aim to design a novel recommendation system that can provide users with image collections relevant to individual personal preferences and interests. To this end, two key issues need to be addressed, i.e., image collection modeling and similarity measurement. For image collection modeling, we consider each image collection as a whole in a group sparse reconstruction framework and extract concise collection descriptors given the pretrained dictionaries. We then consider image collection recommendation as a dynamic similarity measurement problem in response to user's clicked image set, and employ a metric learner to measure the similarity between the image collection and the clicked image set. As there is no previous work directly comparable to this study, we implement several competitive baselines and related methods for comparison. The evaluations on a large scale Pinterest data set have validated the effectiveness of our proposed methods for modeling and recommending image collections.

Given the tremendous growth of online videos, video thumbnail, as the common visualization form of video content, is becoming increasingly important to influence user's browsing and searching experience. However, conventional methods for video thumbnail selection often fail to produce satisfying results as they ignore the side semantic information (e.g., title, description, and query) associated with the video. As a result, the selected thumbnail cannot always represent video semantics and the click-through rate is adversely affected even when the retrieved videos are relevant. In this paper, we have developed a multi-task deep visual-semantic embedding model, which can automatically select query-dependent video thumbnails according to both visual and side information. Different from most existing methods, the proposed approach employs the deep visual-semantic embedding model to directly compute the similarity between the query and video thumbnails by mapping them into a common latent semantic space, where even unseen query-thumbnail pairs can be correctly matched. In particular, we train the embedding model by exploring the large-scale and freely accessible click-through video and image data, as well as employing a multi-task learning strategy to holistically exploit the query-thumbnail relevance from these two highly related datasets. Finally, a thumbnail is selected by fusing both the representative and query relevance scores. The evaluations on 1,000 query-thumbnail dataset labeled by 191 workers in Amazon Mechanical Turk have demonstrated the effectiveness of our proposed method.

In many real-world applications, we are often facing the problem of cross domain learning, i.e., to borrow the labeled data or transfer the already learnt knowledge from a source domain to a target domain. However, simply applying existing source data or knowledge may even hurt the performance, especially when the data distribution in the source and target domain is quite different, or there are very few labeled data available in the target domain. This paper proposes a novel domain adaptation framework, named Semi-supervised Domain Adaptation with Subspace Learning (SDASL), which jointly explores invariant low-dimensional structures across domains to correct data distribution mismatch and leverages available unlabeled target examples to exploit the underlying intrinsic information in the target domain. Specifically, SDASL conducts the learning by simultaneously minimizing the classification error, preserving the structure within and across domains, and restricting similarity defined on unlabeled target examples. Encouraging results are reported for two challenging domain transfer tasks (including image-to-image and image-to-video transfers) on several standard datasets in the context of both image object recognition and video concept detection.

The development of deep learning has empowered machines with comparable capability of recognizing limited image categories to human beings. However, most existing approaches heavily rely on human-curated training data, which hinders the scalability to large and unlabeled vocabularies in image tagging. In this paper, we propose a weakly-supervised deep learning model which can be trained from the readily available Web images to relax the dependence on human labors and scale up to arbitrary tags (categories). Specifically, based on the assumption that features of true samples in a category tend to be similar and noises tend to be variant, we embed the feature map of the last deep layer into a new affinity representation, and further minimize the discrepancy between the affinity representation and its low-rank approximation. The discrepancy is finally transformed into the objective function to give relevance feedback to back propagation. Experiments show that we can achieve a performance gain of 14.0% in terms of a semantic-based relevance metric in image tagging with 63,043 tags from the WordNet, against the typical deep model trained on the ImageNet 1,000 vocabulary set.

One of the fundamental problems in image search is to learn the ranking functions, i.e., similarity between the query and image. The research on this topic has evolved through two paradigms: feature-based vector model and image ranker learning. The former relies on the image surrounding texts, while the latter learns a ranker based on human labeled query-image pairs. Each of the paradigms has its own limitation. The vector model is sensitive to the quality of text descriptions, and the learning paradigm is difficult to be scaled up as human labeling is always too expensive to obtain. We demonstrate in this paper that the above two limitations can be well mitigated by jointly exploring subspace learning and the use of click-through data. Specifically, we propose a novel Ranking Canonical Correlation Analysis (RCCA) for learning query and image similarities. RCCA initially finds a common subspace between query and image views by maximizing their correlations, and further simultaneously learns a bilinear query-image similarity function and adjusts the subspace to preserve the preference relations implicit in the click-through data. Once the subspace is finalized, query-image similarity can be computed by the bilinear similarity function on their mappings in this subspace. On a large-scale click-based image dataset with 11.7 million queries and one million images, RCCA is shown to be powerful for image search with superior performance over several state-of-the-art methods on both keyword-based and query-by-example tasks.

The recommendation of trending images has become a popular feature used by commercial search engines to attract public attention. By browsing through trending images, search engine users can discover trending events at a glance. However, the selection of trending images is very challenging and remains an open issue. Most existing work is highly dependent on editorial efforts, though some preliminarily identify a few plain features for trending images. In this paper, we investigate a set of perceptual factors that can distinguish trending images from common ones. We propose a set of trending-aware features based on several common criteria, which reflect the characteristics of trending images. We further construct a manually labeled dataset based on a commercial search engine's query log over a two-week timespan. We evaluate our proposed method on this dataset and the results demonstrate its effectiveness.

In this article, we present a novel indexing technique called EMIF (Effective Music Indexing Framework) to facilitate scalable and accurate content based music retrieval. It is designed based on a "classification-and-indexing" principle and consists of two main functionality layers: 1) a novel semantic-sensitive classification to identify input music's category and 2) multiple indexing structures - one local indexing structure corresponds to one semantic category. EMIF's layered architecture not only enables superior search accuracy but also reduces query response time significantly. To evaluate the system, a set of comprehensive experimental studies have been carried out using large test collection and EMIF demonstrates promising performance over state-of-the-art approaches.

Similarity search is one of the fundamental problems for large scale multimedia applications. Hashing techniques, as one popular strategy, have been intensively investigated owing to the speed and memory efficiency. Recent research has shown that leveraging supervised information can lead to high quality hashing. However, most existing supervised methods learn hashing function by treating each training example equally while ignoring the different semantic degree related to the label, i.e. semantic confidence, of different examples. In this paper, we propose a novel semi-supervised hashing framework by leveraging semantic confidence. Specifically, a confidence factor is first assigned to each example by neighbor voting and click count in the scenarios with label and click-through data, respectively. Then, the factor is incorporated into the pairwise and triplet relationship learning for hashing. Furthermore, the two learnt relationships are seamlessly encoded into semi-supervised hashing methods with pairwise and listwise supervision respectively, which are formulated as minimizing empirical error on the labeled data while maximizing the variance of hash bits or minimizing quantization loss over both the labeled and unlabeled data. In addition, the kernelized variant of semi-supervised hashing is also presented. We have conducted experiments on both CIFAR-10 (with label) and Clickture (with click data) image benchmarks (up to one million image examples), demonstrating that our approaches outperform the state-of-the-art hashing techniques.

The advent of mobile devices and media cloud services has led to the unprecedented growing of personal photo collections. One of the fundamental problems in managing the increasing number of photos is automatic image tagging. Existing research has predominantly focused on tagging general Web images with a well-labelled image database, e.g., ImageNet. However, they can only achieve limited success on personal photos due to the domain gaps between personal photos and Web images. These gaps originate from the differences in semantic distribution and visual appearance. To deal with these challenges, in this paper, we present a novel transfer deep learning approach to tag personal photos. Specifically, to solve the semantic distribution gap, we have designed an ontology consisting of a hierarchical vocabulary tailored for personal photos. This ontology is mined from $10,000$ active users in Flickr with 20 million photos and 2.7 million unique tags. To deal with the visual appearance gap, we discover the intermediate image representations and ontology priors by deep learning with bottom-up and top-down transfers across two domains, where Web images are the source domain and personal photos are the target. Moreover, we present two modes (single and batch-modes) in tagging and find that the batch-mode is highly effective to tag photo collections. We conducted personal photo tagging on 7,000 real personal photos and personal photo search on the MIT-Adobe FiveK photo dataset. The proposed tagging approach is able to achieve a performance gain of $12.8\%$ and $4.5\%$ in terms of [email protected], against the state-of-the-art hand-crafted feature-based and deep learning-based methods, respectively.

Image tag refinement is the task of refining initial tags of an image such that the refined tags can better reflect the content of the image and, therefore, can help users better access that image. The quality of tag refinement depends on the quality of concept representations that build a mapping from concepts to visual images. While good progress was made in the past decade on tag refinement, the previous approaches only achieved a limited success due to their limited concept representations. In this paper, we show that the visual appearances of a concept consist of both a generic view and a specific view, and therefore we can comprehensively represent a concept by two components. To ensure a clean concept representation, this representation is learned on clean click-through data, where noises are greatly reduced. In the framework, a coarse-to-fine image tag refinement is proposed, which: (1) first generates an efficient star graph to find candidate tags but missing in the initial tag list of an input image and (2) guided by this view-dependent concept representation, formulates a probabilistic objective function to eliminate irrelevant tags. Extensive experiments on two widely used standard data sets (MIRFlickr-25K and NUS-WIDE-270K) demonstrate the effectiveness of our approach.

Exploratory search is emerging as a new form of information-seeking activity in the research community, which generally combines browsing and searching content together to help users gain additional knowledge and form accurate queries, thereby assisting the users with their seeking and investigation activities. However, there have been few attempts at addressing integrated exploratory search solutions when image browsing is incorporated into the exploring loop. In this paper, we investigate the challenges of understanding users' search interests from the product images being browsed and inferring their actual search intentions. We propose a novel interactive image exploring system for allowing users to lightly switch between browse and search processes, and naturally complete visual-based exploratory search tasks in an effective and efficient way. This system enables users to specify their visual search interests in product images by circling any visual objects in web pages, and then the system automatically infers users' underlying intent by analyzing the browsing context and by analyzing the same or similar product images obtained by large-scale image search technology. Users can then utilize the recommended queries to complete intent-specific exploratory tasks. The proposed solution is one of the first attempts to understand users' interests for a visual-based exploratory product search task by integrating the browse and search activities. We have evaluated our system performance based on five million product images. The evaluation study demonstrates that the proposed system provides accurate intent-driven search results and fast response to exploratory search demands compared with the conventional image search methods, and also, provides users with robust results to satisfy their exploring experience.

Landmark summarization with diverse viewpoints is very important in landmark retrieval, as it can create a comprehensive description of a landmark for users. In this paper we present an approach for summarizing a collection of landmark images from diverse viewpoints. First, we group landmark images with content overlap by viewpoint album (VA) generation. Second, we model the relative viewpoint of each image within the VA based on the spatial layout of distinctive descriptors of a landmark. Third, we express the relative viewpoint of an image with a 4-D viewpoint vector, including horizontal, vertical, scale, and rotation. Finally, we summarize the landmarks in terms of viewpoints. Experimental results show the effectiveness of the proposed landmark summarization approach.

With the proliferation of mobile devices, people are taking videos of the same events anytime and anywhere. Even though these crowdsourced videos are uploaded to the cloud and shared, the viewing experience is very limited due to monotonous viewing, visual redundancy, and bad audio-video quality. In this paper, we present a fully automatic mobile video mashup system that works in the cloud to combine recordings captured by multiple devices from different view angles and at different time slots into a single yet enriched and professional looking video-audio stream. We summarize a set of computational filming principles for multicamera settings from a formal focus study. Based on these principles, given a set of recordings of the same event, our system is able to synchronize these recordings with audio fingerprints, assess audio and video quality, detect video cut points, and generate video and audio mashups. The audio mashup is the maximization of audio quality under the less switching principle, while the video mashup is formalized as maximizing video quality and content diversity, constrained by the summarized filming principles. Our system is different from any existing work in this field in three ways: 1) our system is fully automatic; 2) the system incorporates a set of computational domain-specific filming principles summarized from a formal focus study; and 3) in addition to video, we also consider audio mashup that is a key factor of user experience (UX) yet often overlooked in existing research. Evaluations show that our system achieves performance results that are superior to state-of-the-art video mashup techniques, thus providing a better UX.

Leveraging the regularities of people's trajectories, mobility prediction can help forecast social interaction opportunities. In this paper, in order to facilitate real-world social interaction, we aim to predict “serendipitous” social interactions, which are defined as unplanned encounters and interaction opportunities and regarded as emerging social interactions. We collected GPS trajectory data from people' daily life on campus and use it as empirical mobility traces to generate decision trees and model trees to predict next venues, arrival times, and user encounter. Mobility regularities are mainly considered in these prediction models, and mobility contexts (e.g., time, location, and speed) act as decision nodes in the classification trees. Experimental results using collected GPS data showed that our system achieves 90% accuracy for predicting a user's next venue using a decision tree algorithm, with minute-level (around 5 min) prediction error for arrival time using the model tree algorithm. Two prototype applications were developed to support serendipitous social interaction on campus, and the feedback from a user study with 25 users demonstrated the usability of these two applications.

While on the go, people are using their phones as a personal concierge discovering what is around and deciding what to do. Mobile phone has become a recommendation terminal customized for individuals—capable of recommending activities and simplifying the accomplishment of related tasks. In this article, we conduct usage mining on the check-in data, with summarized statistics identifying the local recommendation challenges of huge solution space, sparse available data, and complicated user intent, and discovered observations to motivate the hierarchical, contextual, and sequential solution. We present a point-of-interest (POI) category-transition--based approach, with a goal of estimating the visiting probability of a series of successive POIs conditioned on current user context and sensor context. A mobile local recommendation demo application is deployed. The objective and subjective evaluations validate the effectiveness in providing mobile users both accurate recommendation and favorable user experience.

Techniques for recognizing high-level events in consumer videos on the Internet have many applications. Systems that produced state-of-the-art recognition performance usually contain modules requiring extensive computation, such as the extraction of the temporal motion trajectories, which cannot be deployed on large-scale datasets. In this paper, we provide a comprehensive study on efficient methods in this area and identify technical options for super fast event recognition in Internet videos. We start from analyzing a multimodal baseline that has produced good performance on popular benchmarks, by systematically evaluating each component in terms of both computational cost and contribution to recognition accuracy. After that, we identify alternative features, classifiers, and fusion strategies that can all be efficiently computed. In addition, we also provide a study on the following interesting question: for event recognition in Internet videos, what is the minimum number of visual and audio frames needed to obtain a comparable accuracy to that of using all the frames? Results on two rigorously designed datasets indicate that similar results can be maintained by using only a small portion of the visual frames. We also find that, different from the visual frames, the soundtracks contain little redundant information and thus sampling is always harmful. Integrating all the findings, our suggested recognition system is 2,350-fold faster than a baseline approach with even higher recognition accuracies. It recognizes 20 classes on a 120-second video sequence in just 1.78 seconds, using a regular desktop computer.

From social media has emerged continuous needs for automatic travel recommendations. Collaborative filtering (CF) is the most well-known approach. However, existing approaches generally suffer from various weaknesses. For example , sparsity can significantly degrade the performance of traditional CF. If a user only visits very few locations, accurate similar user identification becomes very challenging due to lack of sufficient information for effective inference. Moreover, existing recommendation approaches often ignore rich user information like textual descriptions of photos which can reflect users' travel preferences. The topic model (TM) method is an effective way to solve the “sparsity problem,” but is still far from satisfactory. In this paper, an author topic model-based collaborative filtering (ATCF) method is proposed to facilitate comprehensive points of interest (POIs) recommendations for social users. In our approach, user preference topics, such as cultural, cityscape, or landmark, are extracted from the geo-tag constrained textual description of photos via the author topic model instead of only from the geo-tags (GPS locations). Advantages and superior performance of our approach are demonstrated by extensive experiments on a large collection of data.

The automatic assessment of photo quality from an aesthetic perspective is a very challenging problem. Most existing research has predominantly focused on the learning of a universal aesthetic model based on hand-crafted visual descriptors . However, this research paradigm can achieve only limited success because (1) such hand-crafted descriptors cannot well preserve abstract aesthetic properties , and (2) such a universal model cannot always capture the full diversity of visual content. To address these challenges, we propose in this paper a novel query-dependent aesthetic model with deep learning for photo quality assessment. In our method, deep aesthetic abstractions are discovered from massive images , whereas the aesthetic assessment model is learned in a query- dependent manner. Our work addresses the first problem by learning mid-level aesthetic feature abstractions via powerful deep convolutional neural networks to automatically capture the underlying aesthetic characteristics of the massive training images . Regarding the second problem, because photographers tend to employ different rules of photography for capturing different images , the aesthetic model should also be query- dependent . Specifically, given an image to be assessed, we first identify which aesthetic model should be applied for this particular image. Then, we build a unique aesthetic model of this type to assess its aesthetic quality. We conducted extensive experiments on two large-scale datasets and demonstrated that the proposed query-dependent model equipped with learned deep aesthetic abstractions significantly and consistently outperforms state-of-the-art hand-crafted feature -based and universal model-based methods.

Current image search engines suffer from a radical variance in retrieval performance over different queries. It is therefore desirable to identify those “difficult” queries in order to handle them properly. Query difficulty estimation is an attempt to predict the performance of the search results returned by an image search system. Most existing methods for query difficulty estimation focus on investigating statistical characteristics of the returned images only, while neglecting very important information , i.e., the query and its relationship with returned images. This relationship plays a crucial role in query difficulty estimation and should be explored further. In this paper we propose a novel query difficulty estimation method with query reconstruction error. This method is proposed based on the observation that, given the images returned for an unknown query, we can easily deduce what the query is from those images if the search results are high quality (i.e., lots of relevant images returned); otherwise, it is difficult to deduce the original query. Therefore, we propose to predict the query difficulty by measuring to what extent the original query can be recovered from the image search results. Specifically, we first reconstruct a visual query from the returned images to summarize their visual theme, and then use the reconstruction error, i.e., the distance between the original textual query and the reconstructed visual query, to estimate the query difficulty. We conduct extensive experiments on two real-world Web image datasets and demonstrate the effectiveness of the proposed method.

The newly emerging event-based social networks (EBSNs) connect online and offline social interactions, offering a great opportunity to understand behaviors in the cyber-physical space. While existing efforts have mainly focused on investigating user behaviors in traditional social network services (SNS), this paper aims to exploit individual behaviors in EBSNs, which remains an unsolved problem. In particular, our method predicts activity attendance by discovering a set of factors that connect the physical and cyber spaces and influence individual's attendance of activities in EBSNs. These factors, including content preference, context (spatial and temporal) and social influence, are extracted using different models and techniques. We further propose a novel Singular Value Decomposition with Multi-Factor Neighborhood (SVD-MFN) algorithm to predict activity attendance by integrating the discovered heterogeneous factors into a single framework, in which these factors are fused through a neighborhood set. Experiments based on real-world data from Douban Events demonstrate that the proposed SVD-MFN algorithm outperforms the state-of-the-art prediction methods.

Social networks nowadays have become an important form of communication in which users can post their current status or share their lives by mobile phones or the Web. In this paper, we develop an effective and efficient model to estimate continuous tie strength between users for friend recommendation with the heterogeneous data from social media community. We categorize those multimodal data into two classes: interaction data (e.g., comments, marking favorite photos) and similarity data(e.g., common friends, groups, tags, geo, visual). We propose to use asymmetric relationship in the interaction data for tie strength estimation instead of using the conventional symmetric ones. Furthermore, by exploring the behavior of users in a social media community, we find that the tie strength between users can be approximately modeled as a linear function of their social connections. Based on this observation, we propose an effective and highly efficient user specific linear model for the tie strength estimation. The experiments on a popular social network show promising results and demonstrate the effectiveness of our proposed method.

Query difficulty estimation (QDE) attempts to automatically predict the performance of the search results returned for a given query. QDE has been widely investigated in text document retrieval for many years. However, few research works have been explored in image retrieval. State-of-the-art QDE methods in image retrieval mainly investigate the statistical characteristics (coherence, robustness, etc.) of the returned images to derive a value for indicating the query difficulty degree. To the best of our knowledge, little research has been done to directly estimate the real retrieval performance of the search results, such as average precision, instead of only an indicator. In this paper, we propose a novel query difficulty estimation approach which automatically estimate the average precision of the image search results. Specifically, we first select a set of query relevant and query irrelevant images for each query via pseudo relevance feedback. Then an efficient and effective voting scheme is proposed to estimate the relevance label of each image in the search results. Based on the images' relevance labels, the average precision of the search results returned for the given query is derived. The experimental results on a benchmark image search dataset demonstrate the effectiveness of the proposed method.

We introduce and investigate a novel problem of image recommendation for web search engine users. Modern web search engines have become a critical assistant for people's daily life. Through interacting with web search engines, users exhibit personalized information needs in various aspects. While this information is critical to improve user experience, it is mostly used only in the web search domain. In this paper, we propose to leverage web search engine users' behavior data to perform image recommendation. To this end, we have developed a two-stage method to label users' preferences for images through crowdsourcing techniques. The two-stage annotation consists of 1) inferring a user's general interests and 2) estimating if this user will be interested in an image. In addition, we implement a baseline algorithm to demonstrate the promise of the proposed cross-domain recommendation framework.

Mobile visual search is attracting much research attention recently. Existing works focus on addressing the limited capacity of wireless channel yet overlook its instability, thus is not adaptive to the change of channel capacity. In this paper, a novel image retrieval algorithm that is scalable to various channel condition is proposed. The proposed algorithm contains three contributions: (1) to achieve instant retrieval under various channel capacity, we adjust transmission load by sparseness instead of codebook size; (2) we introduce hierarchical sparse coding into our retrieval workflow, where original codebook is transformed into a tree-structured dictionary which implies elements' priority; (3) we propose transmission priority ranking schemes that is adaptive to specific query. Experiment results show that the proposed algorithm outperforms BoW and Lasso based algorithm under different parameter settings. Retrieval results under different channel limitation validate the scalability of our method.

Image search engines have achieved good performance for head (popular) queries by leveraging text information and user click data. However, there still remain a large number of tail (rare) queries with relatively unsatisfying search results, which are often overlooked in existing research. Image search for these tail queries therefore provides a grand challenge for research communities. Most existing re-ranking approaches, though effective for head queries, cannot be extended to tail. The assumption of these approaches that the re-ranked list should not go far away from the initial ranked list is not applicable to the tail queries. The challenge, thus, relies on how to leverage the possibly unsatisfying initial ranked results and the very limited click data to solve the search intent gap of tail queries.

In recent years, location-aware music recommendation is increasing in popularity, as more and more users consume music on the move. In this demonstration, we present an intelligent system, called Just-for-Me, to facilitate accurate music recommendation based on where user presents. Our system is developed based on a novel probabilistic generative model, which can effectively integrate the location contexts and global music popularity trends. This approach allows us to gain more comprehensive modeling on user preference and thus significantly enhances the music recommendation performance.

One of the fundamental problems in image search is to rank image documents according to a given textual query. Existing search engines highly depend on surrounding texts for ranking images, or leverage the query-image pairs annotated by human labelers to train a series of ranking functions. However, there are two major limitations: 1) the surrounding texts are often noisy or too few to accurately describe the image content, and 2) the human annotations are resourcefully expensive and thus cannot be scaled up. We demonstrate in this paper that the above two fundamental challenges can be mitigated by jointly exploring the cross-view learning and the use of click-through data. The former aims to create a latent subspace with the ability in comparing information from the original incomparable views (i.e., textual and visual views), while the latter explores the largely available and freely accessible click-through data (i.e., ``crowdsourced" human intelligence) for understanding query. Specifically, we propose a novel cross-view learning method for image search, named Click-through-based Cross-view Learning (CCL), by jointly minimizing the distance between the mappings of query and image in the latent subspace and preserving the inherent structure in each original space. On a large-scale click-based image dataset, CCL achieves the improvement over Support Vector Machine-based method by 4.0\% in terms of relevance, while reducing the feature dimension by several orders of magnitude (e.g., from thousands to tens). Moreover, the experiments also demonstrate the superior performance of CCL to several state-of-the-art subspace learning techniques.

Trending search suggestion is leading a new paradigm of image search, where user's exploratory search experience is facilitated with the automatic suggestion of trending queries. Existing image search engines, however, only provide general suggestions and hence cannot capture user's personal interest. In this paper, we move one step forward to investigate personalized suggestion of trending image searches according to users' search behaviors. To this end, we propose a learning-based framework including two novel components. The first component, i.e., trending-aware weight-regularized matrix factorization (TA-WRMF), is able to suggest personalized trending search queries by learning user preference from many users as well as auxiliary common searches. The second component associates the most representative and trending image with each suggested query. The personalized suggestion of image search consists of a trending textual query and its associated trending image. The combined textual-visual queries not only are trending (bursty) and personalized to user's search preference, but also provide the compelling visual aspect of these queries. We evaluate our proposed learning-based framework on a large-scale search logs with 21 million users and 41 million queries in two weeks from a commercial image search engine. The evaluations demonstrate that our system achieve about 50% gain compared with state-of-the-art in terms of query prediction accuracy.

Auto face annotation, which aims to detect human faces from a facial image and assign them proper human names, is a fundamental research problem and beneficial to many real-world applications. In this work, we address this problem by investigating a retrieval-based annotation scheme of mining massive web facial images that are freely available over the Internet. In particular, given a facial image, we first retrieve the top $(n)$ similar instances from a large-scale web facial image database using content-based image retrieval techniques, and then use their labels for auto annotation. Such a scheme has two major challenges: 1) how to retrieve the similar facial images that truly match the query, and 2) how to exploit the noisy labels of the top similar facial images, which may be incorrect or incomplete due to the nature of web images. In this paper, we propose an effective Weak Label Regularized Local Coordinate Coding (WLRLCC) technique, which exploits the principle of local coordinate coding by learning sparse features, and employs the idea of graph-based weak label regularization to enhance the weak labels of the similar facial images. An efficient optimization algorithm is proposed to solve the WLRLCC problem. Moreover, an effective sparse reconstruction scheme is developed to perform the face annotation task. We conduct extensive empirical studies on several web facial image databases to evaluate the proposed WLRLCC algorithm from different aspects. The experimental results validate its efficacy. We share the two constructed databases "WDB" (714,454 images of 6,025 people) and "ADB" (126,070 images of 1,200 people) with the public. To further improve the efficiency and scalability, we also propose an offline approximation scheme (AWLRLCC) which generally maintains comparable results but significantly reduces the annotation time.

We have witnessed the popularity of image-sharing websites for sharing personal experiences through photos on the Web. These websites allow users describing the content of their uploaded images with a set of tags. Those user-annotated tags are often noisy and biased. Social image tagging aims at removing noisy tags and suggests new relevant tags. However, most existing tag enrichment approaches predominantly focus on tag relevance and overlook tag diversity problem. How to make the top-ranked tags covering a wide range of semantic is still an opening, yet challenging, issue. In this paper, we propose an approach to retag social images with diverse semantics. Both the relevance of a tag to image as well as its semantic compensations to the already determined tags are fused to determine the final tag list for a given image. Different from existing image tagging approaches, the top-ranked tags are not only highly relevant to the image but also have significant semantic compensations with each other. Experiments show the effectiveness of the proposed approach.

Our goal is to boost text-based image search results via image reranking. There are diverse modalities (features) of images that we can leverage for reranking, however, the effects of different modalities are query-dependent. The primary challenge we face is how to fuse multiple modalities adaptively for different queries, which has often been overlooked in previous reranking research. Moreover, multimodality fusion without an understanding of the query is risky, and may lead to incorrect judgment in reranking. Therefore, to obtain the best fusion weights for the query, in this paper, we leverage click-through data, which can be viewed as an "implicit" user feedback and an effective means of understanding the query. A novel reranking algorithm, called click-based relevance feedback, is proposed. This algorithm emphasizes the successful use of click-through data for identifying user search intention, while leveraging multiple kernel learning algorithm to adaptively learn the query-dependent fusion weights for multiple modalities. We conduct experiments on a real-world data set collected from a commercial search engine with click-through data. Encouraging experimental results demonstrate that our proposed reranking approach can significantly improve the NDCG@10 of the initial search results by 11.62%, and can outperform several existing approaches for most kinds of queries, such as tail, middle, and top queries.

The pervasive usage and reach of social media have attracted a surge of attention in the multimedia research community. Community discovery from social media has therefore become an important yet challenging issue. However, due to the subjective generating process, the explicitly observed communities (e.g., group-user and user-user relationship) are often noisy and incomplete in nature. This paper presents a novel approach to discovering communities from social media, including the group membership and user friend structure, by exploring a low-rank matrix recovery technique. In particular, we take Flickr as one exemplary social media platform. We first model the observed indicator matrix of the Flickr community as a summation of a low-rank true matrix and a sparse error matrix. We then formulate an optimization problem by regularizing the true matrix to coincide with the available rich context and content (i.e., photos and their associated tags). An iterative algorithm is developed to recover the true community indicator matrix. The proposed approach leads to a variety of social applications, including community visualization, interest group refinement, friend suggestion, and influential user identification. The evaluations on a large-scale testbed, consisting of 4,919 Flickr users, 1,467 interest groups, and over five million photos, show that our approach opens a new yet effective perspective to solve social network problems with sparse learning technique. Despite being focused on Flickr, our technique can be applied in any other social media community.

With the advent and popularity of social network, more and more users like to share their experiences, such as ratings, reviews, and blogs. The new factors of social network like interpersonal influence and interest based on circles of friends bring opportunities and challenges for recommender system (RS) to solve the cold start and sparsity problem of datasets. Some of the social factors have been used in RS, but have not been fully considered. In this paper, three social factors, personal interest, interpersonal interest similarity, and interpersonal influence, fuse into a unified personalized recommendation model based on probabilistic matrix factorization. The factor of personal interest can make the RS recommend items to meet users' individualities, especially for experienced users. Moreover, for cold start users, the interpersonal interest similarity and interpersonal influence can enhance the intrinsic link among features in the latent space. We conduct a series of experiments on three rating datasets: Yelp, MovieLens, and Douban Movie. Experimental results show the proposed approach outperforms the existing RS approaches.

The four papers in this special section cover several hot and emerging topics in socio-mobile media analysis and retrieval.

The ability to predict when a video search query is not likely to deliver satisfying search results is expected to enable more effective search results optimizations and improved search experience for users. In this paper, we propose a novel context-aware query failure prediction approach that predicts whether a particular query submitted in a user's search session is likely to fail. The approach builds on the well-known concept of query performance prediction introduced in conventional text-based Web search to estimate the query's retrieval performance, but extends this concept with two novel characteristics, user indicators and engine indicators. User indicators are derived from transaction logs, capture the patterns of user interactions with the video search engine, and exploit the context in which a particular query was submitted. Engine indicators are derived from the search results list and measure the consistency of visual search results at the level of visual concepts and textual metadata associated with videos. Extensive evaluation of the approach on a test set containing over one million video search queries shows its effectiveness and demonstrates a significant improvement over traditional and state-of-the-art baseline approaches.

The proliferation of mobile devices is producing a new wave of applications that enable users to sense their surroundings with smart phones. People are preferring mobile devices to search and browse video content on the move. In this paper, we have developed an innovative mobile video search system through which users can discover videos by simply pointing their phones at a screen to capture a very few seconds of what they are watching. Different than most existing mobile video search applications, the proposed system is aiming at instant and progressive video search by leveraging the light-weight computing capacity of mobile devices. In particular, the system is able to index large-scale video data using a new layered audio-video indexing approach in the cloud, as well as generate lightweight joint audio-video signatures with progressive transmission and perform progressive search on mobile devices. Furthermore, we showcase that the system can be applied to two novel applications—video entity search and video clip localization. The evaluations on the real-world mobile video query dataset show that our system significantly improves user’s search experience due to search accuracy, low retrieval latency, and very short recording duration.

Video summarization helps users obtain quick comprehension of video content. Recently, some studies have utilized local features to represent each video frame and formulate video summarization as a coverage problem of local features. However, the importance of individual local features has not been exploited. In this paper, we propose a novel Bag-of-Importance (BoI) model for static video summarization by identifying the frames with important local features as keyframes, which is one of the first studies formulating video summarization at local feature level, instead of at global feature level. That is, by representing each frame with local features, a video is characterized with a bag of local features weighted with individual importance scores and the frames with more important local features are more representative, where the representativeness of each frame is the aggregation of the weighted importance of the local features contained in the frame. In addition, we propose to learn a transformation from a raw local feature to a more powerful sparse nonlinear representation for deriving the importance score of each local feature, rather than directly utilize the hand-crafted visual features like most of the existing approaches. Specifically, we first employ locality-constrained linear coding (LCC) to project each local feature into a sparse transformed space. LCC is able to take advantage of the manifold geometric structure of the high dimensional feature space and form the manifold of the low dimensional transformed space with the coordinates of a set of anchor points. Then we calculate the l2 norm of each anchor point as the importance score of each local feature which is projected to the anchor point. Finally, the distribution of the importance scores of all the local features in a video is obtained as the BoI representation of the video. We further differentiate the importance of local features with a spatial weighting template by taking the perceptual difference among spatial regions of a frame into account. As a result, our proposed video summarization approach is able to exploit both the inter-frame and intra-frame properties of feature representations and identify keyframes capturing both the dominant content and discriminative details within a video. Experimental results on three video datasets across various genres demonstrate that the proposed approach clearly outperforms several state-of-the-art methods.

The popularity of mobile devices equipped with various cameras has revolutionized modern photography. People are able to take photos and share their experiences anytime and anywhere. However, taking a high quality photograph via mobile device remains a challenge for mobile users. In this paper we investigate a photography model to assist mobile users in capturing high quality photos by using both the rich context available from mobile devices and crowdsourced social media on the Web. The photography model is learned from community-contributed images on the Web, and dependent on user's social context. The context includes user's current geo-location, time (i.e., time of the day), and weather (e.g., clear, cloudy, foggy, etc.). Given a wide view of scene, our socialized mobile photography system is able to suggest the optimal view enclosure (composition) and appropriate camera parameters (aperture, ISO, and exposure time). Extensive experiments have been performed for eight well-known hot spot landmark locations where sufficient context tagged photos can be obtained. Through both objective and subjective evaluations, we show that the proposed socialized mobile photography system can indeed effectively suggest proper composition and camera parameters to help the user capture high quality photos.

With the development of image search technology, users are no longer satisfied with searching for images using just metadata and textual descriptions. Instead, more search demands are focused on retrieving images based on similarities in their contents (textures, colors, shapes etc.). Nevertheless, one image may deliver rich or complex content and multiple interests. Sometimes users do not sufficiently define or describe their seeking demands for images even when general search interests appear, owing to a lack of specific knowledge to express their intents. A new form of information seeking activity, referred to as exploratory search, is emerging in the research community, which generally combines browsing and searching content together to help users gain additional knowledge and form accurate queries, thereby assisting the users with their seeking and investigation activities. However, there have been few attempts at addressing integrated exploratory search solutions when image browsing is incorporated into the exploring loop. In this work, we investigate the challenges of understanding users' search interests from the images being browsed and infer their actual search intentions. We develop a novel system to explore an effective and efficient way for allowing users to seamlessly switch between browse and search processes, and naturally complete visual-based exploratory search tasks. The system, called Browse-to-Search enables users to specify their visual search interests by circling any visual objects in the webpages being browsed, and then the system automatically forms the visual entities to represent users' underlying intent. One visual entity is not limited by the original image content, but also encapsulated by the textual-based browsing context and the associated heterogeneous attributes. We use large-scale image search technology to find the associated textual attributes from the repository. Users can then utilize the encapsulated visual entities to complete search tasks. The Browse-to-Search system is one of the first attempts to integrate browse and search activities for a visual-based exploratory search, which is characterized by four unique properties: (1) in session—searching is performed during browsing session and search results naturally accompany with browsing content; (2) in context—the pages being browsed provide text-based contextual cues for searching; (3) in focus—users can focus on the visual content of interest without worrying about the difficulties of query formulation, and visual entities will be automatically formed; and (4) intuitiveness—a touch and visual search-based user interface provides a natural user experience. We deploy the Browse-to-Search system on tablet devices and evaluate the system performance using millions of images. We demonstrate that it is effective and efficient in facilitating the user's exploratory search compared to the conventional image search methods and, more importantly, provides users with more robust results to satisfy their exploring experience.

The rapid growth of the number of videos on the Internet provides enormous potential for users to find content of interest. However, the vast quantity of videos also turns the finding process into a difficult task. In this article, we address the problem of providing personalized video recommendation for users. Rather than only exploring the user-video bipartite graph that is formulated using click information, we first combine the clicks and queries information to build a tripartite graph. In the tripartite graph, the query nodes act as bridges to connect user nodes and video nodes. Then, to further enrich the connections between users and videos, three subgraphs between the same kinds of nodes are added to the tripartite graph by exploring content-based information (video tags and textual queries). We propose an iterative propagation algorithm over the enhanced graph to compute the preference information of each user. Experiments conducted on a dataset with 1,369 users, 8,765 queries, and 17,712 videos collected from a commercial video search engine demonstrate the effectiveness of the proposed method.

Taking the shoe as a concrete example, we present an innovative product retrieval system that leverages object detection and retrieval techniques to support a brand-new online shopping experience in this article. The system, called Circle & Search, enables users to naturally indicate any preferred product by simply circling the product in images as the visual query, and then returns visually and semantically similar products to the users. The system is characterized by introducing attributes in both the detection and retrieval of the shoe. Specifically, we first develop an attribute-aware part-based shoe detection model. By maintaining the consistency between shoe parts and attributes, this shoe detector has the ability to model high-order relations between parts and thus the detection performance can be enhanced. Meanwhile, the attributes of this detected shoe can also be predicted as the semantic relations between parts. Based on the result of shoe detection, the system ranks all the shoes in the repository using an attribute refinement retrieval model that takes advantage of query-specific information and attribute correlation to provide an accurate and robust shoe retrieval. To evaluate this retrieval system, we build a large dataset with 17,151 shoe images, in which each shoe is annotated with 10 shoe attributes e.g., heel height, heel shape, sole shape, etc.). According to the experimental result and the user study, our Circle & Search system achieves promising shoe retrieval performance and thus significantly improves the users' online shopping experience.

The emergence of various and disparate social media platforms has opened opportunities for the research on cross-platform media analysis. This provides huge potentials to solve many challenging problems which cannot be well explored in one single platform. In this paper, we investigate into cross-platform social relation and behavior information to address the cold-start friend recommendation problem. In particular, we conduct an in-depth data analysis to examine what information can better transfer from one platform to another and the result demonstrates a strong correlation for the bidirectional relation and common contact behavior between our test platforms. Inspired by the observations, we design a random walk-based method to employ and integrate these convinced social information to boost friend recommendation performance. To validate the effectiveness of our cross-platform social transfer learning, we have collected a cross-platform dataset including 3,000 users with recognized accounts in both Flickr and Twitter. We demonstrate the effectiveness of the proposed friend transfer methods by promising results.

Traveling experience sharing has become pervasive. In this paper, we present a system which automatically generates a multimedia travelogue for mobile users. Multimedia travelogue shows user footprint with photos on the corresponding location on the map, which could offer the space distribution inner or between the locations. Meanwhile, location overviews with representative tags and images show comprehensive knowledge of the landmarks. The key challenges are: 1) when mapping footprint, some photos are without geo-tags; 2)user-contributed photos may be not clear and graceful; 3) to detect which landmark is actually on the photo to offer location overview. We solve these challenges by combining both geo-tags and tags to estimate the location. First we use the relationship of timestamp and geo-tag in the same trip group to estimate the footprints. Second we localize the landmark on the photo by tag match method which matching user tags with representative tags of the landmark. Experimental results on a Flickr image collection of nearly 2 million images of 6,581 users demonstrate the effectiveness of our approach.

Image search reranking has attracted extensive attention. However, existing image reranking approaches deal with different features independently while ignoring the latent topics among them. It is important to mine multi-latent topic from the features to solve the image search reranking problem. In this paper, we propose a new image reranking model, named reranking with multi-latent topical graph (RMTG), which not only exploits the explicit information of local and global features, but also mines multi-latent topic from these features. We evaluate RMTG over the MSRA-MM dataset and show that RMTG outperforms several existing reranking methods.

Mobile video is quickly becoming a mass consumer phenomenon. More and more people are using their smartphones to search and browse video content while on the move. In this paper, we have developed an innovative instant mobile video search system through which users can discover videos by simply pointing their phones at a screen to capture a very few seconds of what they are watching. The system is able to index large-scale video data using a new layered audio-video indexing approach in the cloud, as well as extract light-weight joint audio-video signatures in real time and perform progressive search on mobile devices. Unlike most existing mobile video search applications that simply send the original video query to the cloud, the proposed mobile system is one of the first attempts at instant and progressive video search leveraging the light-weight computing capacity of mobile devices. The system is characterized by four unique properties: 1) a joint audio-video signature to deal with the large aural and visual variances associated with the query video captured by the mobile phone, 2) layered audio-video indexing to holistically exploit the complementary nature of audio and video signals, 3) light-weight fingerprinting to comply with mobile processing capacity, and 4) a progressive query process to significantly reduce computational costs and improve the user experience---the search process can stop anytime once a confident result is achieved. We have collected 1,400 query videos captured by 25 mobile users from a dataset of 600 hours of video. The experiments show that our system outperforms state-of-the-art methods by achieving 90.79% precision when the query video is less than 10 seconds and 70.07% even when the query video is less than 5 seconds.

This demonstration presents an innovative instant mobile video search system based on layered audio-video indexing, called "LAVES." Through the system, users can discover videos by simply pointing their phones at a screen to capture a very few seconds of what they are watching. Unlike most existing mobile video search applications which simply send the original video query to the cloud, the proposed mobile system is one of the first attempts towards instant and progressive video search leveraging the light-weight computing capacity of mobile devices. The system is able to index large-scale video data using the layered audio-video indexing technique on the cloud, as well as extract light-weight joint audio-video signatures in real time and perform bipartite-graph-based progressive search process on the devices. On a 600 hours video dataset, the system can outperform the state-of-the-arts by achieving 90.79% precision when the query video is less than 10 seconds.

Our objective is to estimate the relevance of an image to a query for image search purposes. We address two limitations of the existing image search engines in this paper. First, there is no straightforward way of bridging the gap between semantic textual queries as well as users' search intents and image visual content. Image search engines therefore primarily rely on static and textual features. Visual features are mainly used to identify potentially useful recurrent patterns or relevant training examples for complementing search by image reranking. Second, image rankers are trained on query-image pairs labeled by human experts, making the annotation intellectually expensive and time-consuming. Furthermore, the labels may be subjective when the queries are ambiguous, resulting in difficulty in predicting the search intention. We demonstrate that the aforementioned two problems can be mitigated by exploring the use of click-through data, which can be viewed as the footprints of user searching behavior, as an effective means of understanding query. The correspondences between an image and a query are determined by whether the image was searched and clicked by users under the query in a commercial image search engine. We therefore hypothesize that the image click counts in response to a query are as their relevance indications. For each new image, our proposed graph-based label propagation algorithm employs neighborhood graph search to find the nearest neighbors on an image similarity graph built up with visual representations from deep neural networks and further aggregates their clicked queries/click counts to get the labels of the new image. We conduct experiments on MSR-Bing Grand Challenge and the results show consistent performance gain over various baselines. In addition, the proposed approach is very efficient, completing annotation of each query-image pair within just 15 milliseconds on a regular PC.

The problem of tagging is mostly considered from the perspectives of machine learning and data-driven philosophy. A fundamental issue that underlies the success of these approaches is the visual similarity, ranging from the nearest neighbor search to manifold learning, to identify similar instances of an example for tag completion. The need to searching for millions of visual examples in high-dimensional feature space, however, makes the task computationally expensive. Moreover, the results can suffer from robustness problem, when the underlying data, such as online videos, are rich of semantics and the similarity is difficult to be learnt from low-level features. This paper studies the exploration of user searching behavior through click-through data, which is largely available and freely accessible by search engines, for learning video relationship and applying the relationship for economic way of annotating online videos. We demonstrated that, by a simple approach using co-click statistics, promising results were obtained in contrast to feature-based similarity measurement. Furthermore, considering the long tail effect that few videos dominate most clicks, a new method based on~polynomial~semantic indexing is proposed to learn a latent space~for alleviating the sparsity problem of click-through data. The proposed approaches are then applied for three major tasks in tagging: tag assignment, ranking, and enrichment. On~a bipartite graph constructed from click-through data with~over 15 million queries and 20 million video URL clicks,~we showed that annotation can be performed for free with competitive performance and minimum computing resource, representing a new and promising paradigm for video tagging in addition to machine learning and data-driven methodologies.

The ongoing revolution in media consumption from traditional PCs to the pervasiveness of mobile devices is driving the adoption of social media in our daily lives. More and more people are using their mobile devices to enjoy social media content while on the move. However, mobile display constraints create challenges for presenting and authoring the rich media content on screens with limited display size. This paper presents an innovative system to automatically generate magazine-like social media visual summaries, which is called "snippet," for efficient mobile browsing. The system excerpts the most salient and dominant elements, i.e., a major picture element and a set of textual elements, from the original media content, and composes these elements into a text overlaid image by maximizing information perception. In particular, we investigate a set of aesthetic rules and visual perception principles to optimize the layout of the extracted elements by considering display constraints. As a result, browsing the snippet on mobile devices is just like quickly glancing at a magazine. To the best of our knowledge, this paper represents one of the first attempts at automatic social media snippet generation by studying aesthetic rules and visual perception principles. We have conducted experiments and user studies with social posts from news entities. We demonstrated that the generated snippets are effective at representing media content in a visually appealing and compact way, leading to a better user experience when consuming social media content on mobile devices.

The search for entities is the most common search behavior on the Web, especially in social media communities where entities (such as images, videos, people, locations, and tags) are highly heterogeneous and correlated. While previous research usually deals with these social media entities separately, we are investigating in this paper a unified, multi-level, and correlative entity graph to represent the unstructured social media data, through which various applications (e.g., friend suggestion, personalized image search, image tagging, etc.) can be realized more effectively in one single framework. We regard the social media objects equally as "entities" and all of these applications as "entity search" problem which searches for entities with different types. We first construct a multi-level graph which organizes the heterogeneous entities into multiple levels, with one type of entities as vertices in each level. The edges between graphs pairwisely connect the entities weighted by intra-relations in the same level and inter-links across two different levels distilled from the social behaviors (e.g., tagging, commenting, and joining communities). To infer the strength of intra-relations, we propose a circular propagation scheme, which reinforces the mutual exchange of information across different entity types in a cyclic manner. Based on the constructed unified graph, we explicitly formulate entity search as a global optimization problem in a unified Bayesian framework, in which various applications are efficiently realized. Empirically, we validate the effectiveness of our unified entity graph for various social media applications on million-scale real-world dataset.

Sign language recognition is a growing research area in the field of computer vision. A challenge within it is to model various signs, varying with time resolution, visual manual appearance, and so on. In this paper, we propose a discriminative exemplar coding (DEC) approach, as well as utilizing Kinect sensor, to model various signs. The proposed DEC method can be summarized as three steps. First, a quantity of class-specific candidate exemplars are learned from sign language videos in each sign category by considering their discrimination. Then, every video of all signs is described as a set of similarities between frames within it and the candidate exemplars. Instead of simply using a heuristic distance measure, the similarities are decided by a set of exemplar-based classifiers through the multiple instance learning, in which a positive (or negative) video is treated as a positive (or negative) bag and those frames similar to the given exemplar in Euclidean space as instances. Finally, we formulate the selection of the most discriminative exemplars into a framework and simultaneously produce a sign video classifier to recognize sign. To evaluate our method, we collect an American sign language dataset, which includes approximately 2000 phrases, while each phrase is captured by Kinect sensor with color, depth, and skeleton information. Experimental results on our dataset demonstrate the feasibility and effectiveness of the proposed approach for sign language recognition.

Search reranking is regarded as a common way to boost retrieval precision. The problem nevertheless is not trivial especially when there are multiple features or modalities to be considered for search, which often happens in image and video retrieval. This paper proposes a new reranking algorithm, named circular reranking, that reinforces the mutual exchange of information across multiple modalities for improving search performance, following the philosophy that strong performing modality could learn from weaker ones, while weak modality does benefit from interacting with stronger ones. Technically, circular reranking conducts multiple runs of random walks through exchanging the ranking scores among different features in a cyclic manner. Unlike the existing techniques, the reranking procedure encourages interaction among modalities to seek a consensus that are useful for reranking. In this paper, we study several properties of circular reranking, including how and which order of information propagation should be configured to fully exploit the potential of modalities for reranking. Encouraging results are reported for both image and video retrieval on Microsoft Research Asia Multimedia image dataset and TREC Video Retrieval Evaluation 2007-2008 datasets, respectively.

Social media has become a very popular way for people to share their photos with friends. Because most of the social images are attached with GPS (geo-tags), a photo's GPS information can be estimated with the help of the large geo-tagged image set while using a visual searching based approach. This paper proposes an unsupervised image GPS location estimation approach with hierarchical global feature clustering and local feature refinement. It consists of two parts: an offline system and an online system. In the offline system, a hierarchical structure is constructed for a large-scale offline social image set with GPS information. Representative images are selected for each GPS location refined cluster, and an inverted file structure is proposed. In the online system, when given an input image, its GPS information can be estimated by hierarchical global clusters selection and local feature refinement in the online system. Both the computational cost and GPS estimation performance demonstrates the effectiveness of the proposed hierarchical structure and inverted file structure in our approach.

This paper describes a novel multimodal interactive image search system on mobile devices. The system, the Joint search with ImaGe, Speech, And Word Plus(JIGSAW+), takes full advantage of the multimodal input and natural user interactions of mobile devices. It is designed for users who already have pictures in their minds but have no precise descriptions or names to address them. By describing it using speech and then refining the recognized query by interactively composing a visual query using exemplary images, the user can easily find the desired images through a few natural multimodal interactions with his/her mobile device. Compared with our previous work JIGSAW, the algorithm has been significantly improved in three aspects: 1) segmentation-based image representation is adopted to remove the artificial block partitions; 2) relative position checking replaces the fixed position penalty; and 3) inverted index is constructed instead of brute force matching. The proposed JIGSAW+ is able to achieve 5% gain in terms of search performance and is ten times faster.

Previous research on online media popularity prediction concluded that the rise in popularity of online videos maintains a conventional logarithmic distribution. However, recent studies have shown that a significant portion of online videos exhibit bursty/sudden rise in popularity, which cannot be accounted for by video domain features alone. In this paper, we propose a novel transfer learning framework that utilizes knowledge from social streams (e.g., Twitter) to grasp sudden popularity bursts in online content. We develop a transfer learning algorithm that can learn topics from social streams allowing us to model the social prominence of video content and improve popularity predictions in the video domain. Our transfer learning framework has the ability to scale with incoming stream of tweets, harnessing physical world event information in real-time. Using data comprising of 10.2 million tweets and 3.5 million YouTube videos, we show that social prominence of the video topic (context) is responsible for the sudden rise in its popularity where social trends have a ripple effect as they spread from the Twitter domain to the video domain. We envision that our cross-domain popularity prediction model will be substantially useful for various media applications that could not be previously solved by traditional multimedia techniques alone.

Mobile devices are becoming ubiquitous. People take pictures via their phone cameras to explore the world on the go. In many cases, they are concerned with the picture-related information. Understanding user intent conveyed by those pictures therefore becomes important. Existing mobile applications employ visual search to connect the captured picture with the physical world. However, they only achieve limited success due to the ambiguity nature of user intent in the picture-one picture usually contains multiple objects. By taking advantage of multitouch interactions on mobile devices, this paper presents a prototype of interactive mobile visual search, named TapTell, to help users formulate their visual intent more conveniently. This kind of search leverages limited yet natural user interactions on the phone to achieve more effective visual search while maintaining a satisfying user experience. We make three contributions in this work. First, we conduct a focus study on the usage patterns and concerned factors for mobile visual search, which in turn leads to the interactive design of expressing visual intent by gesture. Second, we introduce four modes of gesture-based interactions (crop, line, lasso, and tap) and develop a mobile prototype. Third, we perform an in-depth usability evaluation on these different modes, which demonstrates the advantage of interactions and shows that lasso is the most natural and effective interaction mode. We show that TapTell provides a natural user experience to use phone camera and gesture to explore the world. Based on the observation and conclusion, we also suggest some design principles for interactive mobile visual search in the future.

Mobile applications are becoming increasingly popular. More and more people are using their phones to enjoy ubiquitous location-based services (LBS). The increasing popularity of LBS creates a fundamental problem: mobile localization. Besides traditional localization methods that use GPS or wireless signals, using phone-captured images for localization has drawn significant interest from researchers. Photos contain more scene context information than the embedded sensors, leading to a more precise location description. With the goal being to accurately sense real geographic scene contexts, this article presents a novel approach to mobile visual localization according to a given image (typically associated with a rough GPS position). The proposed approach is capable of providing a complete set of more accurate parameters about the scene geo-context including the real locations of both the mobile user and perhaps more importantly the captured scene, as well as the viewing direction. To figure out how to make image localization quick and accurate, we investigate various techniques for large-scale image retrieval and 2D-to-3D matching. Specifically, we first generate scene clusters using joint geo-visual clustering, with each scene being represented by a reconstructed 3D model from a set of images. The 3D models are then indexed using a visual vocabulary tree structure. Taking geo-tags of the database image as prior knowledge, a novel location-based codebook weighting scheme proposed to embed this additional information into the codebook. The discriminative power of the codebook is enhanced, thus leading to better image retrieval performance. The query image is aligned with the models obtained from the image retrieval results, and eventually registered to a real-world map. We evaluate the effectiveness of our approach using several large-scale datasets and achieving estimation accuracy of a user's location within 13 meters, viewing direction within 12 degrees, and viewing distance within 26 meters. Of particular note is our showcase of three novel applications based on localization results: (1) an on-the-spot tour guide, (2) collaborative routing, and (3) a sight-seeing guide. The evaluations through user studies demonstrate that these applications are effective in facilitating the ideal rendezvous for mobile users.

The ever increasing volume of video content on the Web has created profound challenges for developing efficient indexing and search techniques to manage video data. Conventional techniques such as video compression and summarization strive for the two commonly conflicting goals of low storage and high visual and semantic fidelity. With the goal of balancing both video compression and summarization, this article presents a novel approach, called Near-Lossless Semantic Summarization (NLSS), to summarize a video stream with the least high-level semantic information loss by using an extremely small piece of metadata. The summary consists of compressed image and audio streams, as well as the metadata for temporal structure and motion information. Although at a very low compression rate (around 1/40 of H.264 baseline, where traditional compression techniques can hardly preserve an acceptable visual fidelity), the proposed NLSS still can be applied to many video-oriented tasks, such as visualization, indexing and browsing, duplicate detection, concept detection, and so on. We evaluate the NLSS on TRECVID and other video collections, and demonstrate that it is a powerful tool for significantly reducing storage consumption, while keeping high-level semantic fidelity.

Visual reranking has become a widely-accepted method to improve traditional text-based image search results. The main principle is to exploit the visual aggregation property of relevant images among top results so as to boost ranking scores of relevant images, by explicitly or implicitly detecting the confident relevant images, and propagating ranking scores among visually similar images. However, such a visual aggregation property does not always hold, and thus these schemes may fail. In this paper, we instead propose to filter out the most probable irrelevant images using deep contexts, which is the extra information that is not limited in the current search results. The deep contexts for each image consist of sets of images that are returned by searches using the queries formed by the textual context of this image. We compare the popularity of this image in the current search results and the deep contexts to check the irrelevance score. Then the irrelevance scores are propagated to the images whose useful textual context is missed. We formulate the two schemes together to reach a Markov random field, which is effectively solved by graph cuts. The key is that our scheme does not rely on the assumption that relevant images are visually aggregated among top results and is based on the observation that an outlier under the current query is likely to be more popular under some other query. After that, we perform graph reranking over filtered results to reorder them. Experimental results on the INRIA dataset show that our proposed method achieves significant improvements over previous approaches.

While on the go, people are using their phones as a personal concierge discovering what is around and deciding what to do. Mobile phone has become a recommendation terminal customized for individuals. While existing research predominantly focuses on one-step recommendation---recommending the next single activity according to current context, this work moves one step beyond by recommending a series of activities, which is a package of sequential Points of Interest (POIs). The recommended POIs are not only relevant to user context (i.e., current location, time, and check-in), but also personalized to his/her check-in history. We presents a probabilistic approach, which is highly motivated from a large-scale commercial mobile check-in data analysis, to ranking a list of sequential POI categories (e.g., "Japanese food" and "bar") and POIs (e.g., "I love sushi"). The approach enables users to plan consecutive activities on the move. Specifically, the probabilistic recommendation approach estimates the transition probability from one POI to another, conditioned on current context and check-in history in a Markov chain. To alleviate the discritization error and sparsity problem, we further introduce context collaboration and integrate prior information. Experiments on over 100k real-world check-in records and 20k POIs validate the effectiveness of the proposed approach.

Popularity prediction is a key problem in networks to analyze the information diffusion, especially in social media communities. Recently, there have been some custom-build prediction models in Digg and YouTube. However, these models are hardly transplant to an incomplete social network site (e.g., Flickr) by their unique parameters. In addition, because of the large scale of the network in Flickr, it is difficult to get all of the photos and the whole network. Thus, we are seeking for a method which can be used in such incomplete network. Inspired by a collaborative filtering method-Network-based Inference (NBI), we devise a weighted bipartite graph with undetected users and items to represent the resource allocation process in an incomplete network. Instead of image analysis, we propose a modified interdisciplinary models, called Incomplete Network-based Inference (INI). Using the data from 30 months in Flickr, we show the proposed INI is able to increase prediction accuracy by over 58.1%, compared with traditional NBI. We apply our proposed INI approach to personalized advertising application and show that it is more attractive than traditional Flickr advertising.

This paper aims to connect social media from disparate sources on the Internet by building a common topic space in-between, using which cross domain media recommendations can be realized on the web. The topic space is built and updated in real time by extending the Latent Dirichlet Allocation (LDA) model to cater to streaming online data. Our topical model, named Online Streaming LDA (OSLDA), is able to extract, learn, populate, and update the topic space in real time, scaling with streaming tweets. Based on the proposed topic space learned in real time, we present media recommendation applications that cannot be achieved by conventional media analysis techniques: (1) tweet enrichment by recommending related videos, and (2) popular video recommendation for featuring socially trending topical videos. We conduct experiments over a collection of 3.6 million tweets and 1.2 million click-through data from a video search engine. Our results show that the learned topic model plays a natural role connecting cross-domain social media, leading to a better user experience consuming social media.

Capturing a professional photo with high visual quality is always a challenging task for mobile users. This paper presents a crowd sourced learning to photograph approach to assist mobile users for composing high quality photos via their mobile devices. The proposed approach is able to leverage the camera and scene context to search related images with similar context and content from social media communities, and then mine composition knowledge to guide photographing on mobile devices. We develop a patch-based feature generation and selection process to discover salient patches and positions that dominate photo composition aesthetics in the input scene. We then build a regression model to map the composition of salient patches to photo-aesthetic scores. Finally, we develop an efficient hierarchical approach to search for the optimal view enclosure for photograph suggestion. We conducted extensive simulations and subjective evaluations to verify the proposed approach.

The rapid growth of the number of videos on the Internet provides enormous potential for users to find content of interest to them. Video search, such as Google, Youtube, Bing, is a popular way to help users to find desired videos. However, it is still very challenging to discover new video contents for users. In this paper, we address the problem of providing personalized video suggestions for users. Rather than only exploring the user-video graph that is formulated using the click-through information, we also investigate other two useful graphs, the user-query graph indicating if a user ever issues a query, and the query-video graph indicating if a video appears in the search result of a query. The two graphs act as a bridge to connect users and videos, and have a large potential to improve the recommendation as the queries issued by a user essentially imply his interest. As a result, we reach a tripartite graph over (user, video, query). We develop an iterative propagation scheme over the tripartite graph to compute the preference information of each user. Experimental results on a dataset of 2,893 users, 23,630 queries and 55,114 videos collected during Feb. 1-28, 2011 demonstrate that the proposed method outperforms existing state-of-the-art approaches, co-views and random walks on the user-video bipartite graph.

The recent increase in the volume and variety of video content available online presents growing challenges for video search. Users face increased difficulty in formulating effective queries and search engines must deploy highly effective algorithms to provide relevant results. Although lately much effort has been invested in optimizing video search engine results, relatively little attention has been given to predicting for which queries results optimization is most useful, i.e., predicting which queries will fail. Being able to predict when a video search query would fail is likely to make the video search result optimization more efficient and effective, improve the search experience for the user by providing support in the query formulation process and in this way boost the development of video search engines in general. While insight about a query's performance in general could be obtained using the well-known concept of query performance prediction (QPP), we propose a novel approach for predicting a failure of a video search query in the specific context of a search session. Our 'context-aware query failure' prediction approach uses a combination of 'user indicators' and 'engine indicators' to predict whether a particular query is likely to fail in the context of a particular search session. User indicators are derived from the search log and capture the patterns of query (re)formulation behavior and the click-through data of a user during a typical video search session. Engine indicators are derived from the video search results list and capture the visual variance of search results that would be offered to the user for the given query. We validate our approach experimentally on a test set containing 1+ million video search queries and show its effectiveness compared to a set of conventional QPP baselines. Our approach achieves a 13% relative improvement over the baseline.

While on the go, more and more people are using their phones to enjoy ubiquitous location-based services (LBS). One of the fundamental problems of LBS is localization. Researchers are now investigating ways to use a phone-captured image for localization as it contains more scene context information than the embedded sensors. In this paper, we present a novel approach to mobile visual localization that accurately senses geographic scene context according to the current image (typically associated with a rough GPS position). Unlike most existing visual localization methods, the proposed approach is capable of providing a complete set of more accurate parameters about the scene geo---including the actual locations of both the mobile user and perhaps more importantly the captured scene along with the viewing direction. Our approach takes advantage of advanced techniques for large-scale image retrieval and 3D model reconstruction from photos. Specifically, we first perform joint geo-visual clustering in the cloud to generate scene clusters, with each scene represented by a 3D model. The 3D scene models are then indexed using a visual vocabulary tree structure. The phone-captured image is used to retrieve the relevant scene models, then aligned with the models, and further registered to the real-world map. Our approach achieves an estimation accuracy of user location within 14 meters, viewing direction within 9 degrees, and scene location within 21 meters. Such a complete set of accurate geo-parameters can lead to various LBS applications for routing that cannot be achieved with most existing methods. In particular, we showcase three novel applications: 1) accurate self-localization, 2) collaborative localization for rendezvous routing, and 3) routing for photographing. The evaluations through user studies indicate these applications are effective for facilitating the perfect rendezvous for mobile users.

This demonstration presents a novel interactive online shopping application based on visual search technologies. When users want to buy something on a shopping site, they usually have the requirement of looking for related information from other web sites. Therefore users need to switch between the web page being browsed and other websites that provide search results. The proposed application enables users to naturally search products of interest when they browse a web page, and make their even causal purchase intent easily satisfied. The interactive shopping experience is characterized by: 1) in session---it allows users to specify the purchase intent in the browsing session, instead of leaving the current page and navigating to other websites; 2) in context---the browsed web page provides implicit context information which helps infer user purchase preferences; 3) in focus---users easily specify their search interest using gesture on touch devices and do not need to formulate queries in search box; 4) natural-gesture inputs and visual-based search provides users a natural shopping experience. The system is evaluated against a data set consisting of several millions commercial product images.

The usage and applications of social media have become pervasive. This has enabled an innovative paradigm to solve multimedia problems (e.g., recommendation and popularity prediction), which are otherwise hard to address purely by traditional approaches. In this paper, we investigate how to build a mutual connection among the disparate social media on the Internet, using which cross-domain media recommendation can be realized. We accomplish this goal through SocialTransfer---a novel cross-domain real-time transfer learning framework. While existing transfer learning methods do not address how to utilize the real time social streams, our proposed SocialTransfer is able to effectively learn from social streams to help multimedia applications, assuming an intermediate topic space can be built across domains. It is characterized by two key components: 1) a topic space learned in real time from social streams via Online Streaming Latent Dirichlet Allocation (OSLDA), and 2) a real-time cross-domain graph spectra analysis based transfer learning method that seamlessly incorporates learned topic models from social streams into the transfer learning framework. We present as use cases of \emph{SocialTransfer} two video recommendation applications that otherwise can hardly be achieved by conventional media analysis techniques: 1) socialized query suggestion for video search, and 2) socialized video recommendation that features socially trending topical videos. We conduct experiments on a real-world large-scale dataset, including 10.2 million tweets and 5.7 million YouTube videos and show that \emph{SocialTransfer} outperforms traditional learners significantly, and plays a natural and interoperable connection across video and social domains, leading to a wide variety of cross-domain applications.

Mobile visual search has attracted extensive attention for its huge potential for numerous applications. Research on this topic has been focused on two schemes: sending query images, and sending compact descriptors extracted on mobile phones. The first scheme requires about 30-40KB data to transmit, while the second can reduce the bit rate by 10 times. In this paper, we propose a third scheme for extremely low bit rate mobile visual search, which sends compressed visual words consisting of vocabulary tree histogram and descriptor orientations rather than descriptors. This scheme can further reduce the bit rate with few extra computational costs on the client. Specifically, we store a vocabulary tree and extract visual descriptors on the mobile client. A light-weight pre-retrieval is performed to obtain the visited leaf nodes in the vocabulary tree. The orientation of each local descriptor and the tree histogram are then encoded to be transmitted to server. Our new scheme transmits less than 1KB data, which reduces the bit rate in the second scheme by 3 times, and obtains about 30% improvement in terms of search accuracy over the traditional Bag-of-Words baseline. The time cost is only 1.5 secs on the client and 240 msecs on the server.

Mobile devices are ubiquitous. People use their phones as a personal concierge not only discovering information but also searching for particular interest on-the-go and making decisions. This brings a new horizon for multimedia retrieval on mobile. While existing efforts have predominantly focused on understanding textual or a voice query, this paper presents a new perspective which understands visual queries captured by the built-in camera such that mobile-based social activities can be recommended for users to complete. In this work, a query image-based contextual model is proposed for visual search. A mobile user can take a photo and naturally indicate an object-of-interest within the photo via circle based gesture called “O” gesture. Both selected object-of-interest region as well as surrounding visual context in photo are used in achieving a search-based recognition by retrieving similar images based on a large-scale of visual vocabulary tree. Consequently, social activities such as visiting contextually relevant entities (i.e., local businesses) are recommended to the users based on their visual queries and GPS location. Along with the proposed method, an exemplary real application has been developed on Windows Phone 7 devices and evaluated with a wide variety of scenarios on million-scale image database. To test the performance of proposed mobile visual search model, extensive experimentation has been conducted and compared with state-of-the-art algorithms in content-based image retrieval (CBIR) domain.

Automatic photo quality assessment emerged as a hot topic in recent years for its potential in numerous applications. Most existing approaches to photo quality assessment have predominantly focused on image content itself, while ignoring various contexts such as the associated geo-location and timestamp. However, such a universal aesthetic assessment model may not work well with significantly different contexts, since the photography rules are always scene and context dependent. In real cases, professional photographers use different photography knowledge when shooting various scenes in different places. Motivated by this observation, we leverage the geo-context information associated with photos for visual quality assessment. Specifically, we propose in this paper a Scene-Dependent Aesthetic Model (SDAM) to assess photo quality, by jointly leveraging the geo-context and visual content. Geo-contextual leveraged searching is performed to obtain relevant images with similar content to discover the scene-dependent photography principles for accurate photo quality assessment. To overcome the problem that in many cases the number of the contextually searched images is insufficient for learning the SDAM, we adopt transfer learning to utilize auxiliary photos within the same scene category from other locations for learning photography rules. Extensive experiments shows that the proposed SDAM scheme indeed improves the photo quality assessment accuracy via leveraging photo geo-contexts, compared with traditional universal aesthetic models.

The advance of the Internet in the past decade has radically changed the way people communicate and collaborate with each other. Physical distance is no more a barrier in online social networks, but cultural differences (at the individual, community, as well as societal levels) still govern human-human interactions and must be considered and leveraged in the online world. The rapid deployment of high-speed Internet allows humans to interact using a rich set of multimedia data such as texts, pictures, and videos. This position paper proposes to define a new research area called ‘connected multimedia’, which is the study of a collection of research issues of the super-area social media that receive little attention in the literature. By connected multimedia, we mean the study of the social and technical interactions among users, multimedia data, and devices across cultures and explicitly exploiting the cultural differences. We justify why it is necessary to bring attention to this new research area and what benefits of this new research area may bring to the broader scientific research community and the humanity.

The daunting volumes of community-contributed media contents on the Internet have become one of the primary sources for online advertising. However, conventional advertising treats image and video advertising as general text advertising by displaying relevant ads based on the contents of the Web page, without considering the inherent characteristics of visual contents. This article presents a contextual advertising system driven by images, which automatically associates relevant ads with an image rather than the entire text in a Web page and seamlessly inserts the ads in the nonintrusive areas within each individual image. The proposed system, called ImageSense , supports scalable advertising of, from root to node, Web sites, pages, and images. In ImageSense, the ads are selected based on not only textual relevance but also visual similarity, so that the ads yield contextual relevance to both the text in the Web page and the image content. The ad insertion positions are detected based on image salience, as well as face and text detection, to minimize intrusiveness to the user. We evaluate ImageSense on a large-scale real-world images and Web pages, and demonstrate the effectiveness of ImageSense for online image advertising.

Mobile devices are becoming ubiquitous. People use their phones as a personal concierge discovering and making decisions anywhere and anytime. Understanding user intent on the go therefore becomes important for task completion on the phone. While existing efforts have predominantly focused on understanding the explicit user intent expressed by a textual or voice query, this paper presents an approach to context-aware and personalized entity recommendation which understands the implicit intent without any explicit user input on the phone. The approach, highly motivated from a large-scale mobile click-through analysis, is able to rank both the entity types and the entities within each type (here an entity is a local business, e.g., "I love sushi," while an entity type is a category, e.g., "restaurant"). The recommended entity types and entities are relevant to both user context (past behaviors) and sensor context (time and geo-location). Specifically, it estimates the generation probability of an entity by a given user conditioned on the current context in a probabilistic framework. A random-walk propagation is then employed to refine the estimated probability by mining the temporal patterns among entities. We deploy a recommendation application based on the proposed approach on Window Phone 7 devices. We evaluate recommendation performance on 10 thousand mobile clicks, as well as user experience through subjective user studies. We show that the application is effective to facilitate the exploration and discovery of surroundings for mobile users.

This paper proposes a novel audiovisual presentation system, called PhotoSense, to enrich photo navigation experience by associating emotionally harmonic music with a given photo collection. Different from many conventional photo visualization systems which predominantly focus on the visual elements for presentation, we explore both visual and aural perspectives which can enhance the browsing experience from each other. This is achieved by building an emotion space shared by visual and aural domains, and a set of emotion classifiers which can associate each visual and aural element with this space. Furthermore, we design a sequence matching algorithm to associate a set of music with a photo collection by maximizing similarity in the emotion space. Photo-Sense represents one of the first mash-up applications which build a natural connection between the ever increasing personal photo collections on the Web and music-sharing sites. Experiments show that PhotoSense provides better browsing experience for photo collections.

In this paper, we present a novel near-duplicate video retrieval system serving one million web videos. To achieve both the effectiveness and efficiency, a visual word based approach is proposed, which quantizes each video frame into a word and represents the whole video as a bag of words. The system can respond to a query in 41ms with 78.4% MAP on average.

The explosive growth of multimedia data on the Internet creates huge opportunities for multimedia advertising. In this tutorial, we present the techniques and technologies for Internet multimedia advertising. The tutorial aims at bringing together recent insights from the research on multimedia advertising that addresses the theoretical fundamentals, solution concepts, and the issues related to the development of modern multimedia advertising schemes.

The traditional text-based visual search has not been sufficiently improved over the years to accommodate the new emerging demand of mobile users. While on the go, searching on one's phone is becoming pervasive. This paper presents an innovative application for mobile phone users to facilitate their visual search experience. By taking advantage of smart phone functionalities such as multi-modal and multi-touch interactions, users can more conveniently formulate their search intent, and thus search performance can be significantly improved. The system, called JIGSAW (Joint search with ImaGe, Speech, And Words), represents one of the first attempts to create an interactive and multi-modal mobile visual search application. The key of JIGSAW is the composition of an exemplary image query generated from the raw speech via multi-touch user interaction, as well as the visual search based on the exemplary image. Through JIGSAW, users can formulate their search intent in a natural way like playing a jigsaw puzzle on the phone screen: 1) a user speaks a natural sentence as the query, 2) the speech is recognized and transferred to text which is further decomposed to keywords through entity extraction, 3) the user selects preferred exemplary images that can visually represent his/her intent and composes a query image via multi-touch, and 4) the composite image is then used as a visual query to search similar images. We have deployed JIGSAW on a real-world phone system, evaluated the performance on one million images, and demonstrated that it is an effective complement to existing mobile visual search applications.

With the popularity of social media, web users tend to spend more time than before for sharing their experience and interest in online photo-sharing sites. The wide variety of sharing behaviors generate different metadata which pose new opportunities for the discovery of communities. We propose a new approach, named context-based friend suggestion, to leverage the diverse form of contextual cues for more effective friend suggestion in the social media community. Different from existing approaches, we consider both visual and geographical cues, and develop two user-based similarity measurements, i.e., visual similarity and geo similarity for characterizing user relationship. The problem of friend suggestion is casted as a contextual graph modeling problem, where users are nodes and the edges between them are weighted by geo similarity. Meanwhile, the graph is initialized in a way that users with higher visual similarity to a given query have better chance to be recommended. Experimental results on a dataset of 13,876 users and ~1.5 million of their shared photos demonstrated that the proposed approach is consistent with human perception and outperforms other works.

This demonstration presents a mobile-based visual recognition and recommendation application on Windows Phone 7 called TapTell. This is different from other mobile-based visual search mechanisms which merely focus on the search process. TapTell firstly discovers and understands users' visual intents via a circle based natural user interaction called "O" gestures. Following, a Tap action is operated to choose the "O" gestured regions. The context-aware visual search mechanism is utilized for recognizing the intents and associating them with indexed metadata. Finally, the "Tell" action recommends relevant entities utilizing contextual information. The TapTell system has been evaluated at different scenarios on million scale images.

Modeling continuous social strength rather than conventional binary social ties in the social network can lead to a more precise and informative description of social relationship among people. In this paper, we study the problem of social strength modeling (SSM) for the users in a social media community, who are typically associated with diverse form of data. In particular, we take Flickr---the most popular online photo sharing community---as an example, in which users are sharing their experiences through substantial amounts of multimodal contents (e.g., photos, tags, geo-locations, friend lists) and social behaviors (e.g., commenting and joining interest groups). Such heterogeneous data in Flickr bring opportunities yet challenges to the research community for SSM. One of the key issues in SSM is how to effectively explore the heterogeneous data and how to optimally combine them to measure the social strength. In this paper, we present a kernel-based learning to rank framework for inferring the social strength of Flickr users, which involves two learning stages. The first stage employs a kernel target alignment algorithm to integrate the heterogeneous data into a holistic similarity space. With the learned kernel, the second stage rectifies the pair-wise learning to rank approach to estimating the social strength. By learning the social strength graph, we are able to conduct collaborative recommendation and collective classification. The promising results show that the learning-based approach is effective for SSM. Despite being focused on Flickr, our technique can be applied to model social strength of users in any other social media community.

Mobile visual search has been an emerging topic for both researching and industrial communities. Among various methods, visual search has its merit in providing an alternative solution, where text and voice searches are not applicable. This paper proposes an interactive “tap-to-search” approach utilizing both individual's intention in selecting interested regions via “tap” actions on the mobile touch screen, as well as a visual recognition by search mechanism in a large-scale image database. Automatic image segmentation technique is applied in order to provide region candidates. Visual vocabulary tree based search is adopted by incorporating rich contextual information which are collected from mobile sensors. The proposed approach has been conducted on an image dataset with the scale of two million. We demonstrated that using GPS contextual information, such an approach can further achieve satisfactory results with the standard information retrieval evaluation.

Bag-of-words (BOW), which represents an image by the histogram of local patches on the basis of a visual vocabulary, has attracted intensive attention in visual categorization due to its good performance and flexibility. Conventional BOW neglects the contextual relations between local patches due to its Naïve Bayesian assumption. However, it is well known that contextual relations play an important role for human beings to recognize visual categories from their local appearance. This paper proposes a novel contextual bag-of-words (CBOW) representation to model two kinds of typical contextual relations between local patches, i.e., a semantic conceptual relation and a spatial neighboring relation. To model the semantic conceptual relation, visual words are grouped on multiple semantic levels according to the similarity of class distribution induced by them, accordingly local patches are encoded and images are represented. To explore the spatial neighboring relation, an automatic term extraction technique is adopted to measure the confidence that neighboring visual words are relevant. Word groups with high relevance are used and their statistics are incorporated into the BOW representation. Classification is taken using the support vector machine with an efficient kernel to incorporate the relational information. The proposed approach is extensively evaluated on two kinds of visual categorization tasks, i.e., video event and scene categorization. Experimental results demonstrate the importance of contextual relations of local patches and the CBOW shows superior performance to conventional BOW.

Most research on image decomposition, e.g., image segmentation and image parsing, has predominantly focused on the low-level visual clues within a single image and neglected the contextual information across images. In this paper, we present a new perspective to image decomposition piloted by the multilabel context associated with each individual image. Observing that the contextual information (i.e., local label representations of the same label are similar while those from different labels are dissimilar) exists across images, we propose to perform image decomposition in a collective way and obtain an optimal representation for each label from a set of multilabeled images. We formulate the problem as an optimization problem which maximizes inter-label difference while minimizing the intra-label difference of the target label representations and propose two ways to solve this problem. Such a contextual image decomposition has a wide variety of applications, among which two exemplary ones-multilabel image annotation and label ranking, are presented and evaluated with different classification techniques. Extensive experiments on two benchmark datasets demonstrate promising results.

Visual search reranking is defined as reordering visual documents (images or video clips) based on the initial search results or some auxiliary knowledge to improve the search precision. Conventional approaches to visual search reranking empirically take the “classification performance” as the optimization objective, in which each visual document is determined relevant or not, followed by a process of increasing the order of relevant documents. In this paper, we first show that the classification performance fails to produce a globally optimal ranked list, and then we formulate reranking as an optimization problem, in which a ranked list is globally optimal only if any arbitrary two documents in the list are correctly ranked in terms of relevance. This is different from existing approaches which simply classify a document as “relevant” or not. To find the optimal ranked list, we convert the individual documents to “document pairs,” each represented as a “ordinal relation.” Then, we find the optimal document pairs which can maximally preserve the initial rank order while simultaneously keeping the consistency with the auxiliary knowledge mined from query examples and web resources as much as possible. We develop two pairwise reranking methods, difference pairwise reranking (DP-reranking) and exclusion pairwise reranking (EP-reranking), to obtain the relevant relation of each document pair. Finally, a round robin criterion is explored to recover the final ranked list. We conducted comprehensive experiments on an automatic video search task over TRECVID 2005-2007 benchmarks, and showed consistent improvements over text search baseline and other reranking approaches.

With Internet delivery of video content surging to an unprecedented level, video recommendation, which suggests relevant videos to targeted users according to their historical and current viewings or preferences, has become one of most pervasive online video services. This article presents a novel contextual video recommendation system, called VideoReach, based on multimodal content relevance and user feedback. We consider an online video usually consists of different modalities (i.e., visual and audio track, as well as associated texts such as query, keywords, and surrounding text). Therefore, the recommended videos should be relevant to current viewing in terms of multimodal relevance. We also consider that different parts of videos are with different degrees of interest to a user, as well as different features and modalities have different contributions to the overall relevance. As a result, the recommended videos should also be relevant to current users in terms of user feedback (i.e., user click-through). We then design a unified framework for VideoReach which can seamlessly integrate both multimodal relevance and user feedback by relevance feedback and attention fusion. VideoReach represents one of the first attempts toward contextual recommendation driven by video content and user click-through, without assuming a sufficient collection of user profiles available. We conducted experiments over a large-scale real-world video data and reported the effectiveness of VideoReach.

This paper presents an innovative style-wise advertising platform for web page. Web page mainly refers to visual effects, such as color and layout. Unlike the most popular ad-network such as Google AdSense which needs publishers to change the original structure of their pages and define the position and style of the embedded ads manually, style-wise page advertising aims to automatically deliver style-consistent ads at proper positions within the web page, without breaking the layout of the original page. Our system is motivated from the fact that almost 90% web pages contain blank regions without any content. Given a web page with some blank regions, style-wise page advertising is able to detect the suitable blank areas for advertising, rank the ads according to the semantic relevance and web page style, and embed relevant ads into these nonintrusive areas. Style-wise page advertising represents one of the first attempts towards contextual advertising which enables the publishers to save effort when applying online advertising services.

The advent of media-sharing sites has led to the unprecedented Internet delivery of community-contributed media like images and videos. Those visual contents have become the primary sources for online advertising. Conventional advertising treats multimedia advertising as general text advertising by displaying advertisements either relevant to the queries or the Web pages, without considering the potential advantages which could be brought by media contents. In this paper, we summarize the trend of Internet multimedia advertising and conduct a broad survey on the methodologies for advertising which are driven by the rich contents of images and videos. We discuss three key problems in a generic multimedia advertising framework. These problems are:contextual relevancethat determines the selection of relevant advertisements,contextual intrusivenesswhich is the key to detect appropriate ad insertion positions within an image or video, andinsertion optimizationthat achieves the best association between the advertisements and insertion positions so that the effectiveness of advertising can be maximized in terms of both contextual relevance and contextual intrusiveness. We show recently developedMediaSensewhich consists of image, video, and game advertising as an exemplary application of contextual multimedia advertising. In the MediaSense, the most contextually relevant ads are embedded at the most appropriate positions within images or videos. To this end, techniques in computer vision, multimedia retrieval, and computer human interaction are leveraged. We also envision that the next trend of multimedia advertising would be game-like advertising which is more impressionative and thus can promote advertising in an interactive, as well as more compelling and effective way. We conclude this survey with a brief outlook on open research directions.

Most existing approaches to visual search reranking predominantly focus on mining information only from the initial ranking order on the basis of pseudo-relevance feedback. However, the initial ranking order cannot always provide enough cues for reranking by itself due to an unsatisfying visual search performance. This letter presents a novel approach to visual search reranking by selectingtypicalexamples to build the reranking model. Observing that typical examples are mostly clearly visible, fill the majority of the visual documents or appear in one of several common poses, by using these examples informed classifiers would generally be more robust to noisy testing cases that may include occlusions, illumination changes or other factors. We first define thetypicalityon the basis of data distribution, and then theoretically formalize the example selection as an optimization problem on the basis of the example typicality and propose a close-form solution. Based on the selected examples, we build the reranking model by using a support vector machine. Empirically, we conduct extensive experiments on a real-world image set and a benchmark video set, and shows significant and consistent improvements over the state-of-the-art works.

Unlimited vocabulary annotation of multimedia documents remains elusive despite progress solving the problem in the case of a small, fixed lexicon. Taking advantage of the repetitive nature of modern information and online media databases with independent annotation instances, we present an approach to automatically annotate multimedia documents that uses mining techniques to discover new annotations from similar documents and to filter existing incorrect annotations. The annotation set is not limited to words that have training data or for which models have been created. It is limited only by the words in the collective annotation vocabulary of all the database documents. A graph reinforcement method driven by a particular modality (e.g., visual) is used to determine the contribution of a similar document to the annotation target. The graph supplies possible annotations of a different modality (e.g., text) that can be mined for annotations of the target. Experiments are performed using videos crawled from YouTube. A customized precision-recall metric shows that the annotations obtained using the proposed method are superior to those originally existing for the document. These extended, filtered tags are also superior to a state-of-the-art semi-supervised technique for graph reinforcement learning on the initial user-supplied annotations.