Flow Matching (FM) is an efficient generative modeling framework, but aligning it with human preferences remains underexplored.~Although applying Direct Preference Optimization (DPO) to diffusion models has yielded improvements, directly extending DPO-like methods to FM poses three challenges: 1) Incompatibility with ODE-based models, 2) Heavy computational cost from full model fine-tuning, and 3) Reliance on reference model quality. To address these limitations, we propose Preference Classifier for Flow Matching (PC-Flow), a novel reference-free preference alignment framework. Specifically, we reinterpret FM’s deterministic ODE as an equivalent SDE to enable DPO-style learning. Then, we introduce a lightweight classifier to model relative preferences exclusively. This approach decouples alignment from the generative model, eliminating the need for costly fine-tuning or a reference model. Theoretically, PC-Flow guarantees consistent preference-guided distribution evolution, achieves a DPO-equivalent objective without a reference model, and progressively steers generation toward preferred outputs. Experiments show that PC-Flow achieves DPO-level alignment with significantly lower training costs.

Generating high-fidelity full-body human interactions with dynamic objects and static scenes remains a critical challenge in computer graphics and animation. Existing methods for human-object interaction often neglect scene context, leading to implausible penetrations, while human-scene interaction approaches struggle to coordinate fine-grained manipulations with long-range navigation. To address these limitations, we propose HOSIG, a novel framework for synthesizing full-body interactions through hierarchical scene perception. Our method decouples the task into three key components: 1) a scene-aware grasp pose generator that ensures collision-free whole-body postures with precise hand-object contact by integrating local geometry constraints, 2) a heuristic navigation algorithm that autonomously plans obstacle-avoiding paths in complex indoor environments via compressed 2D floor maps and dual-component spatial reasoning, and 3) a scene-guided motion diffusion model that generates trajectory-controlled, full-body motions with finger-level accuracy by incorporating spatial anchors and dual-space gradient-based guidance. Extensive experiments on the TRUMANS dataset demonstrate superior performance over state-of-the-art methods. Notably, our framework supports unlimited motion length through autoregressive generation and requires minimal manual intervention. This work bridges the critical gap between scene-aware navigation and dexterous object manipulation, advancing the frontier of embodied interaction synthesis.

Comprehensively and flexibly capturing the complex spatio-temporal dependencies of human motion is critical for multi-person motion prediction. Existing methods grapple with two primary limitations: i) Inflexible spatiotemporal representation due to reliance on positional encodings for capturing spatiotemporal information. ii) High computational costs stemming from the quadratic time complexity of conventional attention mechanisms. To overcome these limitations, we propose the Spatiotemporal-Untrammelled Mixture of Experts (ST-MoE), which flexibly explores complex spatio-temporal dependencies in human motion and significantly reduces computational cost. To adaptively mine complex spatio-temporal patterns from human motion, our model incorporates four distinct types of spatiotemporal experts, each specializing in capturing different spatial or temporal dependencies. To reduce the potential computational overhead while integrating multiple experts, we introduce bidirectional spatiotemporal Mamba as experts, each sharing bidirectional temporal and spatial Mamba in distinct combinations to achieve model efficiency and parameter economy. Extensive experiments on four multi-person benchmark datasets demonstrate that our approach not only outperforms state-of-art in accuracy but also reduces model parameter by 41.38% and achieves a 3.6× speedup in training.

Document Question Answering (DQA) involves generating answers from a document based on a user's query, representing a key task in document understanding.This task requires interpreting visual layouts, which has prompted recent studies to adopt multimodal Retrieval-Augmented Generation (RAG) that processes page images for answer generation.However, in multimodal RAG, visual DQA struggles to utilize a large number of images effectively, as the retrieval stage often retains only a few candidate pages (e.g., Top-4), causing informative but less visually salient content to be overlooked in favor of common yet low-information pages.To address this issue, we propose a Multi-Armed Bandit-based DQA framework (MAB-DQA) to explicitly model the varying importance of multiple implicit aspects in a query.Specifically, MAB-DQA decomposes a query into aspect-aware subqueries and retrieves an aspect-specific candidate set for each.It treats each subquery as an arm and uses preliminary reasoning results from a small number of representative pages as reward signals to estimate aspect utility.Guided by an exploration-exploitation policy, MAB-DQA dynamically reallocates retrieval budgets toward high-value aspects.With the most informative pages and their correlations, MAB-DQA generates the expected results.On four benchmarks, MAB-DQA shows an average improvement of 5%-18% over the state-of-theart method, consistently enhancing document understanding.

Existing multimodal document question-answering (QA) systems predominantly rely on flat semantic retrieval, representing documents as a set of disconnected text chunks and largely neglecting their intrinsic hierarchical and relational structures. Such flattening disrupts logical and spatial dependencies—such as section organization, figure-text correspondence, and cross-reference relations—that humans naturally exploit for comprehension. To address this limitation, we introduce a document-level structural Document MAP (DMAP), which explicitly encodes both hierarchical organization and inter-element relationships within multimodal documents. Specifically, we design a Structured-Semantic Understanding Agent to construct DMAP by organizing textual content together with figures, tables, charts, etc into a human-aligned hierarchical schema that captures both semantic and layout dependencies. Building upon this representation, a Reflective Reasoning Agent performs structure-aware and evidence-driven reasoning, dynamically assessing the sufficiency of retrieved context and iteratively refining answers through targeted interactions with DMAP. Extensive experiments on MMDocQA benchmarks demonstrate that DMAP yields document-specific structural representations aligned with human interpretive patterns, substantially enhancing retrieval precision, reasoning consistency, and multimodal comprehension over conventional RAG-based approaches. Code is available at https://github.com/Forlorin/DMAP

The key success of existing video super-resolution (VSR) methods stems mainly from exploring spatial and temporal information that is usually achieved by a temporal propagation with alignment strategies. However, inaccurate alignment usually leads to significant artifacts that will be accumulated during propagation and thus affect video restoration. Moreover, only propagating the same timestep features forward or backward does not handle the videos with complex motion or occlusion. To address these issues, we propose a collaborative feedback discriminative (CFD) method to correct inaccurate aligned features and better model spatial and temporal information for VSR. Specifically, we first develop a discriminative alignment correction (DAC) method to reduce the influences of the artifacts caused by inaccurate alignment. Then, we propose a collaborative feedback propagation (CFP) module based on feedback and gating mechanisms to explore spatial and temporal information of different timestep features from forward and backward propagation simultaneously. Finally, we embed the proposed DAC and CFP into commonly used VSR networks to verify the effectiveness of our method. Experimental results demonstrate that our method improves the performance of existing VSR models while maintaining a lower model complexity.

Benefiting from the effectiveness of the self-attention mechanisms in the Transformer framework for modeling non-local features of images, significant progress has been achieved in image super-resolution. We note that existing self-attention mechanisms usually explore all similarities of the tokens between the queries and keys for the feature aggregation. However, using all the similarities does not effectively facilitate the high-quality image reconstruction as not all the tokens from the queries are relevant to those in keys. We further note that self-attention mechanisms are less effective for local feature exploration, which are less effective for the structural detail restoration. To overcome these problems, we develop a simple yet effective adaptive sparse self-attention method to utilize the most useful information of tokens for image restoration. We first develop a local spatial-variant feature estimation method to build the query and key used in the self-attention so that local information can be better modeled. Then, we present a simple yet effective sparse self-attention to adaptively select the most useful similarity values from the self-attention matrix for better the feature aggregation. We analyze that the proposed method models both local and non-local features and thus facilitates better structural detail restoration. We further show that the proposed method can serve as an alternative to existing self-attention mechanisms for better image restoration. Experimental results show that the proposed method performs favorably against state-of-the-art ones on benchmark datasets in terms of accuracy and model complexity.

Few-Shot Action Recognition (FSAR) is a challenging task that requires recognizing novel action categories with a few labeled videos. Recent works typically apply semantically coarse category names as auxiliary contexts to guide the learning of discriminative visual features. However, such context provided by the action names is too limited to provide sufficient background knowledge for capturing novel spatial and temporal concepts in actions. In this paper, we propose DiST, an innovative Decomposition-incorporation framework for FSAR that makes use of decoupled Spatial and Temporal knowledge provided by large language models to learn expressive multi-granularity prototypes. In the decomposition stage, we decouple vanilla action names into diverse spatio-temporal attribute descriptions (action-related knowledge). Such commonsense knowledge complements semantic contexts from spatial and temporal perspectives. In the incorporation stage, we propose Spatial/Temporal Knowledge Compensators (SKC/TKC) to discover discriminative object-level and frame-level prototypes, respectively. In SKC, object-level prototypes adaptively aggregate important patch tokens under the guidance of spatial knowledge. Moreover, in TKC, frame-level prototypes utilize temporal attributes to assist in inter-frame temporal relation modeling. These learned prototypes thus provide transparency in capturing fine-grained spatial details and diverse temporal patterns. Experimental results show DiST achieves state-of-the-art results on five standard FSAR datasets.

Label noise is pervasive in various real-world scenarios, posing challenges in supervised deep learning. Deep networks are vulnerable to such label-corrupted samples due to the memorization effect. One major stream of previous methods concentrates on identifying clean data for training. However, these methods often neglect imbalances in label noise across different mini-batches and devote insufficient attention to out-of-distribution noisy data. To this end, we propose a noise-robust method named Jo-SNC (Joint sample selection and model regularization based on Self- and Neighbor-Consistency). Specifically, we propose to employ the Jensen-Shannon divergence to measure the "likelihood" of a sample being clean or out-of-distribution. This process factors in the nearest neighbors of each sample to reinforce the reliability of clean sample identification. We design a self-adaptive, data-driven thresholding scheme to adjust per-class selection thresholds. While clean samples undergo conventional training, detected in-distribution and out-of-distribution noisy samples are trained following partial label learning and negative learning, respectively. Finally, we advance the model performance further by proposing a triplet consistency regularization that promotes self-prediction consistency, neighbor-prediction consistency, and feature consistency. Extensive experiments on various benchmark datasets and comprehensive ablation studies demonstrate the effectiveness and superiority of our approach over existing state-of-the-art methods.

While text-to-image diffusion models exhibit outstanding results, they struggle to faithfully generate key subjects with corresponding attributes in prompts, challenges known as catastrophic neglect and attribute binding. Previous works typically utilize attention adjustments to solve the above problems, whereas we observe that they may still generate unfaithful images. In this paper, we carefully analyze the text-to-image process and pinpoint three pivotal bottlenecks that hinder image faithful generation: (1) unequal responses of neglected subjects in text embedding, (2) competition and entanglement between subjects' attention, and (3) suboptimal quality of intermediate features from U-Net. Based on the aforementioned observations, we propose a Refine, Control, and Distill (RCD) framework built upon the stable diffusion model to alleviate the negative effects raised by the bottlenecks mentioned above, respectively. Specifically, we achieve the above goals through a text embedding refinement module, three region-level attention control losses, and self-distillation of intermediate semantic features in the denoising process. Our approach exhibits promising capability in generating faithful and high-quality images and outperforms state-of-the-art methods through extensive quantitative and qualitative evaluations on recent advanced base diffusion models.

Fine-grained visual recognition refers to the ability to distinguish subtle differences between visually similar objects— a fundamental yet challenging capability for Multimodal Large Language Models (MLLMs). In this paper, we observe that even strong open-source MLLMs, such as Qwen2-VL and InternVL2, still struggle with accurately identifying fine-grained categories. These models often fail to attend to subtle but critical details for precise discrimination. To unlock this potential, we propose FineG-RAG, a retrieval-augmented generation pipeline designed to enhance the fine-grained recognition capabilities of MLLMs. FineG-RAG integrates external fine-grained knowledge into the recognition process via a generalized retriever. To support this, we construct fine-grained visual-language knowledge database containing representative images with wide visual diversity and expert-crafted attribute descriptions from multiple perspectives. Relevant fine-grained knowledge is retrieved from this database and fed into a visual-language augmented prompt, which provides rich multimodal context to guide MLLMs in generating accurate labels. To better evaluate the fine-grained recognition capabilities of MLLMs, we design a multiple-choice evaluation strategy based on publicly four fine-grained datasets. Extensive experiments demonstrate that FineG-RAG consistently outperforms baseline methods, achieving superior recognition accuracy across a range of off-the-shelf, open-source MLLMs.

Although memory-based video colorization methods can effectively leverage information from previous frames to guide the colorization of the current frame, they still struggle to capture the semantic features of grayscale frames. To overcome this limitation, we introduce a foundation model (Dinov3 [1]) with strong semantic priors to distill semantic features from grayscale frames, thereby improving the accuracy of feature retrieval from memory and enhancing the effect of subsequent object matching. Moreover, existing memory-based approaches often suffer from redundant information, which can degrade colorization quality. To address this issue, we design a dynamic feature selection module (DFSM) that adaptively selects the most relevant features from memory. In addition, we propose a multi-scale fusion module (MSFM) that fuses mid- and high-scale features derived from visual–semantic interactions to further refine video colorization. To minimize the loss of useful information, we discard the feature-compression strategy adopted in previous methods, thereby preserving temporal features to the greatest extent and further improving colorization performance. Extensive experiments on benchmark datasets and real-world videos demonstrate that the proposed method outperforms existing state-of-the-art approaches, validating its effectiveness and performance advantages.

The growing demand for Ultra-High-Definition (UHD) resolution imaging challenges the efficiency of image enhancement techniques. However, existing low-light enhancement methods exhibit a strong correlation between computational workload and image resolution, resulting in significant slowdowns in inference speed when handling UHD images. To overcome this problem, we present an effective and efficient approach for low-light image enhancement, named Sparse Curve Estimation (SCE). In contrast to curve-based methods that use quadratic curves for pixel-wise refinement, we propose a dual control curve formulation and estimate sparse curves for the entire image, which can be converted into Look-Up Tables (LUTs). Then, the iterative curve computation is replaced by a simple indexing and color-spatial interpolation operation from the LUTs, thereby accelerating inference speed. With a resized input, our method requires minimal operations, regardless of the size of the image. Furthermore, we develop a histogram smoothness loss function to equalize the histograms of the images. Quantitative and qualitative results demonstrate the effectiveness of the proposed approach. Our approach outperforms the state-of-the-art in terms of inference speed; specifically, it is 11× faster than existing approaches in the 4K setting.

Dense-localization Audio-Visual Events (DAVE) aims to identify time boundaries and corresponding categories for events that are both audible and visible in a long video, where events may co-occur and exhibit varying durations. However, complex audio-visual scenes often involve asynchronization between modalities, making accurate localization challenging. Existing DAVE solutions extract audio and visual features through unimodal encoders, and fuse them via dense cross-modal interaction. However, independent unimodal encodingstruggles to emphasize shared semantics between modalitieswithout cross-modal guidance, while dense cross-modal attention mayover-attend to semantically unrelated audio-visual features. To address these problems, we present LOCO, a Locality-aware cross-modal Correspondence learning framework for DAVE. LOCO leverages the local temporal continuity of audio-visual events as important guidance to filter irrelevant cross-modal signals and enhance cross-modal alignment throughout both unimodal and cross-modal encoding stages. i) Specifically, LOCO applies Local Correspondence Feature (LCF) Modulation to enforce unimodal encoders to focus on modality-shared semantics by modulating agreement between audio and visual features based on local cross-modal coherence. ii) To better aggregate cross-modal relevant features, we further customize Local Adaptive Cross-modal (LAC) Interaction, which dynamically adjusts attention regions in a data-driven manner. This adaptive mechanism focuses attention on local event boundaries and accommodates varying event durations. By incorporating LCF and LAC, LOCO provides solid performance gains and outperforms existing DAVE methods. The source code will be released.

Diffusion model-based networks have been widely applied in the field of image generation and have gradually demonstrated a strong potential in image colorization tasks. However, despite the emergence of various colorization diffusion models, two major challenges remain: (1) the lack of effective control over the colorization process and (2) the prevalent issue of color bleeding. Integrating suitable conditional control can effectively alleviate these challenges. To this end, we propose a unified multi-modal diffusion model that harnesses diverse modality information to achieve flexible and high-quality colorization. Specifically, we introduce a Stroke-Adapter that extracts and integrates stroke prompt, enhancing user control over color distribution. Additionally, we design an Edge-Guided Attention mechanism to effectively inject edge information into the colorization process, significantly reducing color bleeding artifacts. Extensive comparative experiments demonstrate that our method outperforms state-of-the-art image colorization approaches in both qualitative and quantitative evaluations, achieving superior colorization results with enhanced controllability.

Skeleton-based human action recognition has emerged as a prominent research focus in computer vision, with significant progress achieved in recent years. However, existing methods often suffer substantial performance degradation under real-world data constraints, such as body occlusion, missing frames, and noise. These limitations critically undermine the robustness of related techniques in practical applications. To address these challenges, we propose a Spatial Temporal Self-compensating Graph Convolutional Network (STSc-GCN), which skillfully utilizes the systematic and regular nature of human movement to mitigate performance degradation caused by data constraints through a data self-compensation mechanism. Specifically, STSc-GCN comprises two key modules: 1) collaborative motion spatial compensation (CMSC). This module designs multiple distinct topological relationships, primarily including Walk-probability Generality Topology and Self-organizing Particularity Topology, respectively, to deeply explore the universal and personalized collaborative relationships between human joints. These relationships help compensate for the lack of information caused by spatial data constraints and 2) meta-action sharpening temporal Compensation (MSTC). This module introduces a novel motion sharpening mechanism that enhances key dynamic information within the meta-action sequences through cross-attention technology, thereby improving model adaptability to missing-frame scenarios. STSc-GCN achieves state-of-the-art performance on four constrained datasets and shows superior results on three widely used standard datasets, confirming its effectiveness in both constrained and general scenarios. Code will be available at https://github.com/XingLi1012/STSc-GCN.git.

Noisy Correspondence (NC), caused by mismatched pairs in multimedia datasets, poses major challenges for cross-modal retrieval, especially under high noise levels. Existing solutions often suffer from substantial performance degradation as noise levels increase. To address this issue, we propose Pseudo-Text guided Robust Learning (PTRL), a novel framework designed to identify noisy pairs and enhance model robustness. Specifically, PTRL leverages pseudo-text as explicit supervision signals and introduces a new data division criterion to accurately distinguish between clean and noisy pairs. Instead of discarding or directly using noisy data, PTRL proposes a pseudo-text replacement strategy to maintain semantic consistency of the training set, thereby facilitating more reliable learning. In addition, pseudo-text-image pairs serve as a form of data augmentation, enriching data diversity and improving model generalization. To further stabilize training and mitigate overfitting, PTRL incorporates a robust InfoNCE loss that is particularly effective in the presence of noise. Extensive experiments demonstrate that PTRL achieves state-of-the-art performance and robustness, with an RSum improvements of +60.1% on Flickr30K and +22.6% on MS-COCO at an 80% noise level, significantly outperforming existing methods. The datasets and source code are available at https://github.com/shidan0122/PTRL.git.

Unsupervised domain adaptation semantic segmentation (UDASS) aims to perform dense prediction on the unlabeled target domain by training the model on a labeled source domain. In this field, self-training approaches have demonstrated strong competitiveness and advantages. However, existing methods often rely on additional training data (such as reference datasets or depth maps) to rectify the unreliable pseudo-labels, ignoring the cross-domain interaction between the target and source domains. To address this issue, in this paper, we propose a novel method for unsupervised domain adaptation semantic segmentation, termed Unlocking Cross-Domain Synergies (UCDS). Specifically, in the UCDS network, we design a new Dynamic Self-Correction (DSC) module that effectively transfers source domain knowledge and generates high-confidence pseudo-labels without additional training resources. Unlike the existing methods, DSC proposes a Dynamic Noisy Label Detection method for the target domain. To correct the noisy pseudo-labels, we design a Dual Bank mechanism that explores the reliable and unreliable predictions of the source domain, and conducts cross-domain synergy through Weighted Reassignment Self-Correction and Negative Correction Prevention strategies. To enhance the discriminative ability of features and amplify the dissimilarity of different categories, we propose Discrepancy-based Contrastive Learning (DCL). The DCL selects positive and negative samples in the source and target domains based on the semantic discrepancies among different categories, effectively avoiding the numerous false negative samples found in existing methods. Extensive experimental results on three commonly used datasets demonstrate the superiority of the proposed UCDS in comparison with the state-of-the-art methods. The project and code are available at https://github.com/wqh011128/UCDS.

Existing weakly supervised hashing often suffers from the imprecision of user-provided tags and over-reliance on textual knowledge from pre-trained word embeddings, neglecting crucial visual knowledge associated with image labels. As a result, this leads to unsatisfactory performance in closed-vocabulary tasks and limited generalization in open-vocabulary scenarios. To address this issue, we propose Multi-modal Knowledge Distillation Hashing (MKDH), a novel method leveraging visual and language pre-training (VLP) model such as CLIP to learn robust hash codes. Our method designs a dual-layer attention adapter to generate joint representations by capturing fine-grained visual and textual knowledge from the CLIP teacher network. Additionally, we introduce a knowledge extraction contrastive loss to enhance the robustness of joint representations and a knowledge distillation contrastive loss to transfer the extracted multi-modal knowledge to the hash codes. To further mitigate the negative impact of false negative pairs in these contrastive losses, we introduce false negative weighting strategy that reduces the weights assigned to such pairs. Extensive experiments on three widely used datasets demonstrate that our method achieves robust retrieval performance with significant improvements in both closed- and open-vocabulary settings. The source code is available athttps://github.com/IMAG-LZY/MKDH.

We have recently seen some progress in the current field of human-human interaction generation. However, directly generating complex two-person interactive motions remains a significant challenge. Meanwhile, these models typically employ two independent timelines when generating motions for interactive scenarios involving two individuals. This design overlooks the temporal dependencies between motions at each timestep and fails to account for the roles of active and reactive participants during the generation process, often resulting in unrealistic and unnatural motions. In this work, we propose HiTMM, a novel framework for Human interaction generation based on Temporal Masked Modeling. HiTMM first decomposes the human interaction into two separate single-person motions. Individual motions within the interaction belong to the same type, enabling them to be mapped to a shared latent space through a coarse-to-fine approach that produces multi-layer discrete tokens. We then arrange all tokens of the two interacting individuals along a shared timeline. Subsequently, we employ a masked transformer and a residual transformer to model the base-layer and rest-layer motion tokens. Both the base-layer and rest-layer motion tokens are arranged along a single timeline, allowing the model to explicitly capture the temporal order and initiating role embedded in the sequence, where the first individual's motion initiates the interaction. Note that, our model utilizes a shared temporal representation, making it capable of performing temporal editing on specific regions within human interaction sequences. Experimental results show that our model achieves an FID of 5.017 on the InterHuman dataset, surpassing the current state-of-the-art model (vs 5.154 for InterMask), and an FID of 0.373 on the InterX dataset (vs 0.399 for InterMask).

This paper presents IMAGGarment, a fine-grained garment generation (FGG) framework that enables high-fidelity garment synthesis with precise control over silhouette, color, and logo placement. Unlike existing methods that are limited to single-condition inputs, IMAGGarment addresses the challenges of multi-conditional controllability in personalized fashion design and digital apparel applications. Specifically, IMAGGarment employs a two-stage training strategy to separately model global appearance and local details, while enabling unified and controllable generation through end-to-end inference. In the first stage, we propose a global appearance model that jointly encodes silhouette and color using a mixed attention module and a color adapter. In the second stage, we present a local enhancement model with an adaptive appearance-aware module to inject user-defined logos and spatial constraints, enabling accurate placement and visual consistency. To support this task, we release GarmentBench, a large-scale dataset comprising over 180 K garment samples paired with multi-level design conditions, including sketches, color references, logo placements, and textual prompts. Extensive experiments demonstrate that our method outperforms existing baselines, achieving superior structural stability, color fidelity, and local controllability performance.

Existing virtual try-on (VTON) methods provide only limited user control over garment attributes and generally overlook essential factors such as face, pose, and scene context. To address these limitations, we introduce the virtual dressing (VD) task, which aims to synthesize freely editable human images conditioned on fixed garments and optional user-defined inputs. We further propose a comprehensive affinity metric index (CAMI) to quantify the consistency between generated outputs and reference garments. We present IMAGDressing-v1, which leverages a garment-specific U-Net to integrate semantic features from CLIP and texture features from a VAE. To incorporate these garment features into a frozen denoising U-Net for flexible text-driven scene control, we employ a hybrid attention mechanism composed of frozen self-attention and trainable cross-attention layers. IMAGDressing-v1 seamlessly integrates with extension modules, such as ControlNet and IP-Adapter, enabling enhanced diversity and controllability. To alleviate data constraints, we introduce the Interactive Garment Pairing (IGPair) dataset, comprising over 300,000 garment–image pairs and a standardized data assembly pipeline. Extensive experiments demonstrate that IMAGDressing-v1 achieves state-of-the-art performance in controlled human image synthesis. The code and model will be available at https://github.com/muzishen/IMAGDressing.

Recent advances in diffusion models focus on efficiently handling conditional generative tasks without extra training. The process involves decomposing the result into two components: 1. unconditional sample, generated in the absence of conditions; 2. condition correction, adjusting unconditional sample to include the guidance image. This adjustment is quantified by the pixel-level measure, where the latent is decoded back into a pixel image, and the forward operator translates the noisy image into the guidance domain for comparison with the guidance image. To enhance the fidelity of condition correction, we propose a learnable latent forward operator, focusing on latent-space consistency with the expectation that this latent-space consistency approximates the pixel-level fidelity measure. The encoder translates the guidance image into the latent space, and a correctional operator is proposed to rectify model mismatching in the latent guidance model. The determination of the condition term and the correction estimation is akin to solving a blind inverse problem. Our EMControl employs the Expectation-Maximization (EM) algorithm to solve the blind inverse problem during the reverse sampling process. This technique ensures that samples, once consistent with the guidance, are accurately mapped back onto the noisy data manifold, adhering to the data's inherent distribution. The EMControl has proven its effectiveness by delivering superior performance in conditional diffusion generation tasks compared to previous approaches. Moreover, its application to multiple-condition scenarios underscores its versatility and robustness across a range of generative tasks.

Given the critical role of birds in ecosystems, Fine-Grained Bird Recognition (FGBR) has gained increasing attention, particularly in distinguishing birds within similar subcategories. Although Vision Transformer (ViT)-based methods often outperform Convolutional Neural Network (CNN)-based methods in FGBR, recent studies reveal that the limited receptive field of plain ViT model hinders representational richness and makes them vulnerable to scale variance. Thus, enhancing the multi-scale capabilities of existing ViT-based models to overcome this bottleneck in FGBR is a worthwhile pursuit. In this paper, we propose a novel framework for FGBR, namely Multi-scale Diverse Cues Modeling (MDCM), which explores diverse cues at different scales across various stages of a multi-scale Vision Transformer (MS-ViT) in an ``Activation-Selection-Aggregation'' paradigm. Specifically, we first propose a multi-scale cue activation module to ensure the discriminative cues learned at different stage are mutually different. Subsequently, a multi-scale token selection mechanism is proposed to remove redundant noise and highlight discriminative, scale-specific cues at each stage. Finally, the selected tokens from each stage are independently utilized for bird recognition, and the recognition results from multiple stages are adaptively fused through a multi-scale dynamic aggregation mechanism for final model decisions. Both qualitative and quantitative results demonstrate the effectiveness of our proposed MDCM, which outperforms CNN- and ViT-based models on several widely-used FGBR benchmarks.

In training-free conditional generative tasks, diffusion models utilize differentiable loss functions to steer the generative reverse process, necessitating modifications to sampling algorithms like DDPM and DDIM. However, such adjustments likely reduce flexibility and reliability. In this paper, we propose NoiseCtrl, a sampling-algorithm-agnostic technique for controlled image generation. Essentially, diffusion models generate denoised results zt−1by adding a predicted meanµtwith random noise ϵt. NoiseCtrl specifically adjusts the random noise while leaving the underlying sampling algorithms unchanged. At each step t, NoiseCtrl converts the unconditional Gaussian noise into conditional noise $\varepsilon _t^\prime $ by substituting the isotropic Gaussian distribution with the von Mises–Fisher distribution. This substitution introduces a directional focus while preserving the randomness required for conditional image generation. Thanks to this non-intrusive design, NoiseCtrl is straightforward to integrate and has been extensively validated through experiments, demonstrating its adaptability for different diffusion algorithms and superior performance across various conditional generation tasks.

Convolutional neural networks (CNNs) and Vision Transformers (ViTs) have achieved excellent performance in image restoration. While ViTs generally outperform CNNs by effectively capturing long-range dependencies and input-specific characteristics, their computational complexity increases quadratically with image resolution. This limitation hampers their practical application in high-resolution image restoration. In this paper, we propose a simple yet effective visual state space model (EVSSM) for image deblurring, leveraging the benefits of state space models (SSMs) for visual data. In contrast to existing methods that employ several fixed-direction scanning for feature extraction, which significantly increases the computational cost, we develop an efficient visual scan block that applies various geometric transformations before each SSM-based module, capturing useful non-local information and maintaining high efficiency. In addition, to more effectively capture and represent local information, we propose an efficient discriminative frequency domain-based feedforward network (EDFFN), which can effectively estimate useful frequency information for latent clear image restoration. Extensive experimental results show that the proposed EVSSM performs favorably against state-of-the-art methods on benchmark datasets and real-world images.

Despite the significant progress made by all-in-one models in universal image restoration, existing methods suffer from a generalization bottleneck in real-world scenarios, as they are mostly trained on small-scale synthetic datasets with limited degradations. Therefore, large-scale high-quality real-world training data is urgently needed to facilitate the emergence of foundational models for image restoration. To advance this field, we spare no effort in contributing a million-scale dataset with two notable advantages over existing training data: real-world samples with larger-scale, and degradation types with higher diversity. By adjusting internal camera settings and external imaging conditions, we can capture aligned image pairs using our well-designed data acquisition system over multiple rounds and our data alignment criterion. Moreover, we propose a robust model, FoundIR, to better address a broader range of restoration tasks in real-world scenarios, taking a further step toward foundation models. Specifically, we first utilize a diffusion-based generalist model to remove degradations by learning the degradation-agnostic common representations from diverse inputs, where incremental learning strategy is adopted to better guide model training. To refine the model's restoration capability in complex scenarios, we introduce degradation-aware specialist models for achieving final high-quality results. Extensive experiments show the value of our dataset and the effectiveness of our method.

Recent advances in conditional diffusion models have shown promise for generating realistic TalkingFace videos, yet challenges persist in achieving consistent head movement, synchronized facial expressions, and accurate lip synchronization over extended generations. To address these, we introduce the Motion-priors Conditional Diffusion Model (MCDM), which utilizes both archived and current clip motion priors to enhance motion prediction and ensure temporal consistency. The model consists of three key elements: (1) an archived-clip motion-prior that incorporates historical frames and a reference frame to preserve identity and context; (2) a present-clip motion-prior diffusion model that captures multimodal causality for accurate predictions of head movements, lip sync, and expressions; and (3) a memory-efficient temporal attention mechanism that mitigates error accumulation by dynamically storing and updating motion features. We also introduce the TalkingFace-Wild dataset, a multilingual collection of over 200 hours of footage across 10 languages. Experimental results demonstrate the effectiveness of MCDM in maintaining identity and motion continuity for long-term TalkingFace generation.

Large vision-language models (LVLMs) have significantly advanced numerous fields. In this work, we explore how to harness their potential to address 3D scene understanding tasks, using 3D question answering (3D-QA) as a representative example. Due to the limited training data in 3D, we do not train LVLMs but infer in a zero-shot manner. Specifically, we sample 2D views from a 3D point cloud and feed them into 2D models to answer a given question. When the 2D model is chosen, e.g., LLAVA-OV, the quality of sampled views matters the most. We propose cdViews, a novel approach to automatically selecting critical and diverse Views for 3D-QA. cdViews consists of two key components: viewSelector prioritizing critical views based on their potential to provide answer-specific information, and viewNMS enhancing diversity by removing redundant views based on spatial overlap. We evaluate cdViews on the widely-used ScanQA and SQA benchmarks, demonstrating that it achieves state-of-the-art performance in 3D-QA while relying solely on 2D models without fine-tuning. These findings support our belief that 2D LVLMs are currently the most effective alternative (of the resource-intensive 3D LVLMs) for addressing 3D tasks.

In current training-free Conditional Diffusion Models (CDM), the sampling process is steered by the gradient, which measures the discrepancy between the guidance and the condition extracted by a pre-trained condition extraction network. These methods necessitate small guidance steps, resulting in longer sampling times. To address the issue of slow sampling, we introduce AccCtr, a method that simplifies the conditional sampling algorithm by maximizing the sum of two objectives. The local maximum set of one objective is contained within the local maximum set of the other. Leveraging this relationship, we decompose the joint optimization into two parts, alternately maximizing each objective. By analyzing the steps involved in optimizing these objectives, we identify the most time-consuming steps and recommend retraining condition extraction network—a relatively simple task—to reduce its computational cost. Integrating AccCtr into current CDMs is a seamless task that does not impose a significant computational burden. Extensive testing has demonstrated that AccCtr offers superior sample quality and faster generation times.

Visible-infrared person re-identification (VIReID) retrieves pedestrian images with the same identity across different modalities. Existing methods learn visual features solely from images, failing to align them into the modality-invariant semantic space. In this paper, we propose a novel framework, termed Richer Semantics, Better Alignment (RSBA), to align visual features with explicit and enriched semantics. Specifically, we first develop an Explicit Semantics-Guided Feature Alignment (ESFA) module, which supplements textual descriptions for cross-modality images and aligns image-text pairs within each modality, alleviating the distribution discrepancy of visual features. We then devise a Consistent Similarity-Guided Indirect Alignment (CSIA) module, which constrains the similarity between intra-modality image-text pairs to be consistent with that between inter-modality text-text pairs, indirectly aligning visual features with cross-modality semantics. Furthermore, we design a Cross-View Semantics Compensation (CVSC) module, which integrates multi-view texts and improves the image-text matching of one-to-one in ESFA and CSIA to one-to-many, further strengthening the alignment of visual features within the semantic space. Extensive experimental results on three public datasets demonstrate the effectiveness and superiority of our proposed RSBA.

Text-to-image person retrieval (TIPR) aims to find images of the same identity that match a given text description. Current TIPR methods mainly focus on mining the association between images and texts, ignoring their potential complementarity. Besides, existing matching losses treat all positive pairs from the same identity equally, leading to noisy correspondences. In this paper, we propose CoRL: a cross-modal Collaborative Representation Learning framework designed to improve TIPR by effectively leveraging the complementarity between modalities. The text typically contains identity details with less noise, which helps distinguish visually similar pedestrians. This inspires us to integrate it into the corresponding image to emphasize identity-related and modality-shared visual information. However, corresponding text for each image is not always available, especially during inference. Accordingly, we introduce a Virtual-text Embedding Synthesizer that generates high-quality virtual-text features for cross-modal collaboration, eliminating the need for actual texts. We then design a Cross-Modal Collaboration learning process, incorporating a Cross-modal Relation Consistency loss to promote interaction and fusion between image and virtual-text features for mutual enhancement. Additionally, an Identity-bounded Matching loss is proposed to handle different types of image-text pairs distinctly, leading to more accurate cross-modal correspondences. Extensive experiments on multiple benchmarks demonstrate the superiority of CoRL over existing TIPR methods.

Recently, adapting Vision Language Models (VLMs) to zero-shot visual classification by tuning class embedding with a few prompts (Test-time Prompt Tuning, TPT) or replacing class names with generated visual samples (support-set) has shown promising results. However, TPT cannot avoid the semantic gap between modalities while the support-set cannot be tuned. To this end, we draw on each other's strengths and propose a novel framework, namely TEst-time Support-set Tuning for zero-shot Video Classification (TEST-V). It first dilates the support-set with multiple prompts (Multi-prompting Support-set Dilation, MSD) and then erodes the support-set via learnable weights to mine key cues dynamically (Temporal-aware Support-set Erosion, TSE). Specifically, i) MSD expands the support samples for each class based on multiple prompts inquired from LLMs to enrich the diversity of the support-set. ii) TSE tunes the support-set with factorized learnable weights according to the temporal prediction consistency in a self-supervised manner to dig pivotal supporting cues for each class. TEST-V achieves state-of-the-art results across four benchmarks and shows good interpretability.

Recent diffusion model advancements aim to handle conditional generative tasks without extra training. Existing training-free methods add a correction term at each denoising step, but they often face computational instability and lack controllability, especially with limited samples and large noise. We propose a new approach using the von Mises-Fisher (vMF) distribution to model the denoised result, turning the conditional generation task into an estimation problem for vMF parameters. We formulate the conditional diffusion model as a mean vector estimation problem for the Gaussian distribution, noting that this can be seen as an estimation problem from noisy observations. When the sampling number is small, the estimation is unstable. To address this, we optimize the mean vector of the vMF distribution by minimizing the KL divergence between the prior and posterior distributions. This approach not only addresses the computational instability but also improves the controllability and quality of the generated results. Once these parameters are determined, the denoised result can be sampled directly from the vMF distribution. Estimating the parameters requires minimal additional code and incurs negligible computational overhead while significantly improving performance. Extensive experiments across various conditional generation tasks, including depth maps, edge detection, segmentation, and style guidance, demonstrate the superiority and versatility of our method. Our approach consistently outperforms existing training-free methods and even surpasses some training-required methods in terms of visual quality and controllability.

Vision-language models (VLMs) has demonstrated impressive cross-modal alignment. However, their internal mechanisms of associating text concepts with visual patterns remain opaque. This opacity raises a critical question: What visual patterns do VLMs inherently associate with text concepts? Current methods for decoding representations of VLMs often produce suboptimal outputs, hindering to probe the clear visual patterns. To address this, we introduce Generative Semantic Probing (GSP), a novel training-free framework that synthesizes images to probe the implicit semantic preferences of VLMs. Our method generates visual patterns that maximize the similarity to the target text embeddings, through three core components: (1) Hierarchical Feature Decomposition, which decomposes the image generation across multi-scale feature levels; (2) Feature Space Constraint, which constrains the optimization within semantically meaningful feature subspace; (3) Quality Assessment Module, which ensures the generation of visually plausible outputs. Experiments validate our method's strengths in high-fidelity image generation and interpretable model analysis. Beyond text-to-image generation, style transfer and image editing applications, our framework enables unprecedented visualization of VLMs' decision boundaries. By exposing implicit preferences and systematic biases in the cross-modal association, our work provides a valuable insight for both understanding and improvement of the vision-language alignment.

Dance is an important art form in human culture, but creating new dances can be both challenging and time-consuming. In this paper, we propose a novel dance choreography framework, EDMG, designed to efficiently generate creative and long-lasting dance sequences conditioning on music and dance descriptions. In the first stage, we propose a flexible dance diffusion method, combined with dance genre description and descriptions of fundamental movements to generate the dance sequences. To achieve high computational efficiency and inference speed, EDMG designs a lightweight denoising module by using selective parallel scanning algorithm from Mamba2. This Parallel Mamba Denoiser reduces significantly the number of parameters and accelerates remarkably both the learning and inference processes. In the second stage, by designing a smoothing module with a long receptive field, we mitigate joint error accumulation that causes jittering movements and foot sliding, thereby enhancing the fluency and visual appeal of the dance movements. Furthermore, we extend the AIST++ dataset by adding detailed descriptions of dance genres and fundamental movements, using the Large Language Model (LLM). These descriptions further improve the choreography generation. EDMG is validated through extensive experiments, demonstrating that our method can both effectively and efficiently generate long-term dances suitable for various dance genres. Project URL: https://github.com/neymar277/EDMG.

Text-to-image diffusion models are typically trained on large-scale web data, often resulting in outputs that misalign with human preferences. Inspired by preference learning in large language models, we propose ABC (Alignment by Classification), a simple yet effective framework for aligning diffusion models with human preferences. In contrast to prior DPO-based methods that depend on suboptimal supervised fine-tuned (SFT) reference models, ABC assumes access to an ideal reference model perfectly aligned with human intent and reformulates alignment as a classification problem. Under this view, we recognize that preference data naturally forms a semi-supervised classification setting. To address this, we propose a data augmentation strategy that transforms preference comparisons into fully supervised training signals. We then introduce a classification-based ABC loss to guide alignment. Our alignment by classification approach could effectively steer the diffusion model toward the behavior of the ideal reference. Experiments on various diffusion models show that our ABC consistently outperforms existing baselines, offering a scalable and robust solution for preference-based text-to-image fine-tuning.

A major challenge in using diffusion models is aligning outputs with user-defined conditions. Existing conditional generation methods fall into two major categories: classifier-based guidance, which requires differentiable target models and gradient-based correction; and classifier-free guidance, which embeds conditions directly into the diffusion model but demands expensive joint training and architectural coupling. In this work, we introduce a third paradigm: DISCrete nOise (DISCO) guidance, which replaces the continuous conditional correction term with a finite codebook of discrete noise vectors sampled from a Gaussian prior. Conditional generation is reformulated as a code selection task, and we train prediction network to choose the optimal code given the intermediate diffusion state and the conditioning input. Our approach is differentiability-free, and training-efficient, avoiding the gradient computation and architectural redundancy of prior methods. Empirical results demonstrate that DISCO achieves competitive controllability while substantially reducing resource demands, positioning it as a scalable and effective alternative for conditional diffusion generation.

Current 3D gaze estimation methods struggle to generalize across diverse data domains, primarily due to $\textbf{i)}$ $\textit{the scarcity of annotated datasets}$, and $\textbf{ii)}$ $\textit{the insufficient diversity of labeled data}$. In this work, we present OmniGaze, a semi-supervised framework for 3D gaze estimation, which utilizes large-scale unlabeled data collected from diverse and unconstrained real-world environments to mitigate domain bias and generalize gaze estimation in the wild. First, we build a diverse collection of unlabeled facial images, varying in facial appearances, background environments, illumination conditions, head poses, and eye occlusions. In order to leverage unlabeled data spanning a broader distribution, OmniGaze adopts a standard pseudo-labeling strategy and devises a reward model to assess the reliability of pseudo labels. Beyond pseudo labels as 3D direction vectors, the reward model also incorporates visual embeddings extracted by an off-the-shelf visual encoder and semantic cues from gaze perspective generated by prompting a Multimodal Large Language Model to compute confidence scores. Then, these scores are utilized to select high-quality pseudo labels and weight them for loss computation. Extensive experiments demonstrate that OmniGaze achieves state-of-the-art performance on five datasets under both in-domain and cross-domain settings. Furthermore, we also evaluate the efficacy of OmniGaze as a scalable data engine for gaze estimation, which exhibits robust zero-shot generalization on four unseen datasets.

We present *Plenodium* (*plenoptic medium*), an effective and efficient 3D representation framework capable of jointly modeling both objects and the participating medium. In contrast to existing medium representations that rely solely on view-dependent modeling, our novel plenoptic medium representation incorporates both directional and positional information through spherical harmonics encoding, enabling highly accurate underwater scene reconstruction. To address the initialization challenge in degraded underwater environments, we propose the pseudo-depth Gaussian complementation to augment COLMAP-derived point clouds with robust depth priors. In addition, a depth ranking regularized loss is developed to optimize the geometry of the scene and improve the ordinal consistency of the depth maps. Extensive experiments on real-world underwater datasets demonstrate that our method achieves significant improvements in 3D reconstruction. Furthermore, we construct a simulated dataset with GT and the controllable scattering medium to demonstrate the restoration capability of our method in underwater scenarios.

Real-world applications are stretching context windows to hundreds of thousand of tokens while Large Language Models (LLMs) swell from billions to trillions of parameters. This dual expansion send compute and memory costs skyrocketing, making $\textit{token compression}$ indispensable. We introduce Vision Centric Token Compression ($\textbf{Vist}$), a $\textit{slow–fast}$ compression framework that mirrors human reading: the $\textit{fast}$ path renders distant tokens into images, letting a $\textbf{frozen, lightweight vision encoder}$ skim the low-salience context; the $\textit{slow}$ path feeds the proximal window into the LLM for fine-grained reasoning. A Probability-Informed Visual Enhancement (PVE) objective masks high-frequency tokens during training, steering the Resampler to concentrate on semantically rich regions—just as skilled reader gloss over function words. On eleven in-context learning benchmarks, $\textbf{Vist}$ achieves the same accuracy with 2.3$\times$ fewer tokens, cutting FLOPs by 16\% and memory by 50\%. This method delivers remarkable results, outperforming the strongest text encoder-based compression method CEPE by $\textbf{7.6}$\% on average over benchmarks like TriviaQA, NQ, PopQA, NLUI, and CLIN, setting a new standard for token efficiency in LLMs. The project is at https://github.com/CSU-JPG/VIST.

Recent years have witnessed significant advances in image deraining due to the progress of effective image priors and deep learning models. As each deraining approach has individual settings (e.g., training and test datasets, evaluation criteria), how to fairly evaluate existing approaches comprehensively is not a trivial task. Although existing surveys aim to thoroughly review image deraining approaches, few of them focus on unifying evaluation settings to examine the deraining capability and practicality evaluation. In this paper, we provide a comprehensive review of existing image deraining methods and provide a unified evaluation setting to evaluate their performance. Furthermore, we construct a new high-quality benchmark named HQ-RAIN to conduct extensive evaluations, consisting of 5,000 paired high-resolution synthetic images with high harmony and realism. We also discuss existing challenges and highlight several future research opportunities worth exploring. To facilitate the reproduction and tracking of the latest deraining technologies for general users, we build an online platform to provide the off-the-shelf toolkit, involving the large-scale performance evaluation.

In this paper, we propose the Vision-Audio-Language Omni-peRception pretraining model (VALOR) for multimodal understanding and generation. Unlike widely-studied vision-language pretraining models, VALOR jointly models the relationships among vision, audio, and language in an end-to-end manner. It consists of three separate encoders for single modality representations and a decoder for multimodal conditional text generation. We design two pretext tasks to pretrain the VALOR model: Multimodal Grouping Alignment (MGA) and Multimodal Grouping Captioning (MGC). MGA projects vision, language, and audio into the same common space, simultaneously building vision-language, audio-language, and audiovisual-language alignment. MGC learns to generate text tokens under conditions of vision, audio, or both. To promote vision-audio-language pretraining research, we construct a large-scale, high-quality tri-modality dataset named VALOR-1M, containing 1 million audible videos with human-annotated audiovisual captions. Extensive experiments show that VALOR can learn strong multimodal correlations and generalize to various downstream tasks (e.g., retrieval, captioning, and question answering) with different input modalities (e.g., vision-language, audio-language, and audiovisual-language). VALOR achieves new state-of-the-art performance on a series of public cross-modality benchmarks.

How to effectively explore spatial and temporal information is important for video deblurring. In contrast to existing methods that directly align adjacent frames without discrimination, we develop a deep discriminative spatial and temporal network to facilitate the spatial and temporal feature exploration for better video deblurring. We first develop a channel-wise gated dynamic network to adaptively explore the spatial information. As adjacent frames usually contain different contents, directly stacking features of adjacent frames without discrimination may affect the latent clear frame restoration. Therefore, we develop a simple yet effective discriminative temporal feature fusion module to obtain useful temporal features for latent frame restoration. Moreover, to utilize the information from long-range frames, we develop a wavelet-based feature propagation method that takes the discriminative temporal feature fusion module as the basic unit to effectively propagate main structures from long-range frames for better video deblurring. Experimental results show that the proposed method performs favorably against state-of-the-art ones on benchmark datasets in terms of accuracy and model complexity.

RGB-Thermal Salient Object Detection (RGB-T SOD) aims to pinpoint prominent objects within aligned pairs of visible and thermal infrared images. A key challenge lies in bridging the inherent disparities between RGB and Thermal modalities for effective saliency map prediction. Traditional encoder-decoder architectures, while designed for cross-modality feature interactions, may not have adequately considered the robustness against noise originating from defective modalities, thereby leading to suboptimal performance in complex scenarios. Inspired by hierarchical human visual systems, we propose the ConTriNet, a robust Confluent Triple-Flow Network employing a "Divide-and-Conquer" strategy. This framework utilizes a unified encoder with specialized decoders, each addressing different subtasks of exploring modality-specific and modality-complementary information for RGB-T SOD, thereby enhancing the final saliency map prediction. Specifically, ConTriNet comprises three flows: two modality-specific flows explore cues from RGB and Thermal modalities, and a third modality-complementary flow integrates cues from both modalities. ConTriNet presents several notable advantages. It incorporates a Modality-induced Feature Modulator (MFM) in the modality-shared union encoder to minimize inter-modality discrepancies and mitigate the impact of defective samples. Additionally, a foundational Residual Atrous Spatial Pyramid Module (RASPM) in the separated flows enlarges the receptive field, allowing for the capture of multi-scale contextual information. Furthermore, a Modality-aware Dynamic Aggregation Module (MDAM) in the modality-complementary flow dynamically aggregates saliency-related cues from both modality-specific flows. Leveraging the proposed parallel triple-flow framework, we further refine saliency maps derived from different flows through a flow-cooperative fusion strategy, yielding a high-quality, full-resolution saliency map for the final prediction. To evaluate the robustness and stability of our approach, we collect a comprehensive RGB-T SOD benchmark, VT-IMAG, covering various real-world challenging scenarios. Extensive experiments on public benchmarks and our VT-IMAG dataset demonstrate that ConTriNet consistently outperforms state-of-the-art competitors in both common and challenging scenarios, even when dealing with incomplete modality data. The code and VT-IMAG will be available at: https://cser-tang-hao.github.io/contrinet.html.

Multi-modal learning aims to enhance performance by unifying models from various modalities but often faces the "modality imbalance" problem in real data, leading to a bias towards dominant modalities and neglecting others, thereby limiting its overall effectiveness. To address this challenge, the core idea is to balance the optimization of each modality to achieve a joint optimum. Existing approaches often employ a modal-level control mechanism for adjusting the update of each modal parameter. However, such a global-wise updating mechanism ignores the different importance of each parameter. Inspired by subnetwork optimization, we explore a uniform sampling-based optimization strategy and find it more effective than global-wise updating. According to the findings, we further propose a novel importance sampling-based, element-wise joint optimization method, called Adaptively Mask Subnetworks Considering Modal Significance (AMSS). Specifically, we incorporate mutual information rates to determine the modal significance and employ non-uniform adaptive sampling to select foreground subnetworks from each modality for parameter updates, thereby rebalancing multi-modal learning. Additionally, we demonstrate the reliability of the AMSS strategy through convergence analysis. Building upon theoretical insights, we further enhance the multi-modal mask subnetwork strategy using unbiased estimation, referred to as AMSS+. Extensive experiments reveal the superiority of our approach over comparison methods.

The discussion of compositional generalization in action recognition,i.e., Compositional Action Recognition (CAR), has recently received increasing attention. CAR challenges models to recognize unseen combinations of actions and objects, with the primary challenge being the distribution shift from training to testing. Most previous approaches for CAR incorporate supplementary object annotations (e.g. bounding boxes and objects categories) to learn an instance-centric dynamic representation. However, these methods inevitably introduce stronger visual inductive bias, including object appearance and background bias, that impact generalization performance, particularly in out-of-distribution scenarios. To this end, this work attempts to construct an appearance-agnostic de-biased representation by leveraging the powerful segmentation capability of Segment Anything Model (SAM), which is the first exploration of SAM in the field of compositional action recognition. Specifically, we propose a novel SAM-driven Appearance-Agnostic Representation Learning (A2RL) framework for CAR, which contains two effective sub-modules: Fore-Back Mask (FBM) and Dynamic Relation Modeling (DRM). In FBM, we design a fine-grained instance-invisible and background-removed masking strategy to effectively weaken the strong connection between visual cues and action labels, as well as minimize the impact of irrelevant factors. In DRM, we explore the potential association between subjects and objects involved in one action and then build appearance-agnostic relational descriptors for dynamic modeling. Extensive experiments demonstrate the generalization ability of this work. Notably, FBM achieves significant improvements in all three compositional settings without adding any additional model parameters. The proposed also gains state-of-the-art performance in comparison with the most recent methods in CAR.

The vision transformer (ViT) architecture offers significant advantages in object detection tasks. However, some limitations affect improving task performance. Firstly, the ViT relies heavily on inflexible position embedding, which causes poor performance when processing images with complex semantic dependencies.Secondly, class imbalance in large-scale datasets can easily cause training instability and inference bias. To overcome these limitations, we propose a generalized concordant ViT scheme for object detection (GCViTDet). Specifically, we first introduce a relevance enhancement strategy (RES) into the encoder-decoder structure, which is composed of the spatial enhanced position embeddings (SEPE) component, the cross multipooling attention (CMPA) component, and a global-local path. This strategy establishes semantic-rich dependencies through enhanced position embedding information and omni-feature representations. Subsequently, a bottom-up feature aggregation pathway is employed, utilizing a cross multi-pooling attention to improve the model’s capacity to capture semantic dependencies. This scheme enables the extraction of high-dimensional features that exhibit complex positional relationships. Besides, we propose a focal unified cross-entropy (FUCE) loss to solve the class imbalance problem during training by introducing a uniform threshold to regulate the similarity between positive and negative samples of different classes. Compared with existing methods, GCViTDet can not only capture more intricate positional relationships and semantic-rich dependencies but also alleviate the class-imbalance problem. Experimental results on the challenging MS-COCO dataset validate that GCViTDet can consistently improve performance over state-of-the-art object detection baseline models.

Understanding how the brain responds to external stimuli and decoding this process has been a significant challenge in neuroscience. While previous studies typically concentrated on brain-to-image and brain-to-language reconstruction, our work strives to reconstruct gestures associated with speech stimuli perceived by brain. Unfortunately, the lack of paired {brain, speech, gesture} data hinders the deployment of deep learning models for this purpose. In this paper, we introduce a novel approach, fMRI2GES, that allows training of fMRI-to-gesture reconstruction networks on unpaired data using Dual Brain Decoding Alignment. This method relies on two key components: (i) observed texts that elicit brain responses, and (ii) textual descriptions associated with the gestures. Then, instead of training models in a completely supervised manner to find a mapping relationship among the three modalities, we harness an fMRIto- text model, a text-to-gesture model with paired data and an fMRI-to-gesture model with unpaired data, establishing dual fMRI-to-gesture reconstruction patterns. Afterward, we explicitly align two outputs and train our model in a self-supervision way. We show that our proposed method can reconstruct expressive gestures directly from fMRI recordings. We also investigate fMRI signals from different ROIs in the cortex and how they affect generation results. Overall, we provide new insights into decoding co-speech gestures, thereby advancing our understanding of neuroscience and cognitive science.

Visible-Infrared Person Re-Identification (VI-ReID) is a challenging task due to the large modality discrepancy between visible and infrared images, which complicates the alignment of their features into a suitable common space. Moreover, style noise, such as illumination and color contrast, reduces the identity discriminability and modality invariance of features. To address these challenges, we propose a novel Diverse Semantics-guided Feature Alignment and Decoupling (DSFAD) network to align identity-relevant features from different modalities into a textual embedding space and disentangle identity-irrelevant features within each modality. Specifically, we develop a Diverse Semantics-guided Feature Alignment (DSFA) module, which generates pedestrian descriptions with diverse sentence structures to guide the cross-modality alignment of visual features. Furthermore, to filter out style information, we propose a Semantic Margin-guided Feature Decoupling (SMFD) module, which decomposes visual features into pedestrian-related and style-related components, and then constrains the similarity between the former and the textual embeddings to be at least a margin higher than that between the latter and the textual embeddings. Additionally, to prevent the loss of pedestrian semantics during feature decoupling, we design a Semantic Consistency-guided Feature Restitution (SCFR) module, which further excavates useful information for identification from the style-related features and restores it back into the pedestrian-related features, and then constrains the similarity between the features after restitution and the textual embeddings to be consistent with that between the features before decoupling and the textual embeddings. Extensive experiments on three VI-ReID datasets demonstrate the superiority of our DSFAD. The code will be made publicly available at https://github.com/nengdong96/DSFAD.

In hashing-based long-tailed image retrieval, the dominance of data-rich head classes often hinders the learning of effective hash codes for data-poor tail classes due to inherent long-tailed bias. Interestingly, this bias also contains valuable prior knowledge by revealing inter-class dependencies, which can be beneficial for hash learning. However, previous methods have not thoroughly analyzed this tangled negative and positive effects of long-tailed bias from a causal inference perspective. In this paper, we propose a novel hash framework that employs causal inference to disentangle detrimental bias effects from beneficial ones. To capture good bias in long-tailed datasets, we construct hash mediators that conserve valuable prior knowledge from class centers. Furthermore, we propose a de-biased hash loss To enhance the beneficial bias effects while mitigating adverse ones, leading to more discriminative hash codes. Specifically, this loss function leverages the beneficial bias captured by hash mediators to support accurate class label prediction, while mitigating harmful bias by blocking its causal path to the hash codes and refining predictions through backdoor adjustment. Extensive experimental results on four widely used datasets demonstrate that the proposed method improves retrieval performance against the state-of-the-art methods by large margins. The source code is available at https://github.com/IMAG-LuJin/CIH.

In this paper, we propose a novel visual relation detection task, named Group Visual Relation Detection (GVRD), for detecting visual relations whose subjects and/or objects are groups (GVRs), inspired by the observation that groups are common in image semantic representation. GVRD can be deemed as an evolution over the existing visual relation detection task that limits both subjects and objects of visual relations as individuals. We propose a Simultaneous Group Relation Prediction (SGRP) method that can simultaneously predict groups and predicates to address GVRD. SGRP contains an Entity Construction (EC) module, a Feature Extraction (FE) module, and a Group Relation Prediction (GRP) module. Specifically, the EC module constructs instances, group candidates, and phrase candidates; the FE module extracts visual, location and semantic features for these entities; and the GRP module simultaneously predicts groups and predicates, and generates the GVRs. Moreover, we construct a new dataset, named COCO-GVR, to facilitate solutions to GVRD task, which consists of 9,570 images from COCO dataset and 31,855 manually labeled GVRs. We test and validate the performance of SGRP by extensive experiments on COCO-GVR dataset. It shows that SGRP outperforms the baselines generated from the state-of-the-art visual relation detection and scene graph generation methods.

Edge sensor devices generate vast amounts of user data, but centralized processing poses privacy risks. Federated Learning addresses this by decentralizing training. However, applying Federated Learning directly to skeleton videos fails to preserve motion dynamics and suffers from client heterogeneity bias. To address these limitations, we propose CSAR-a Client-Unbiased Skeletal Action Recognizer for Federated Learning-which tackles two core challenges: motion dynamics preservation and classifier bias mitigation. Specifically, CSAR employs a Model Calibration Loss during client training to align client-server representations and reduce drift. On the server, it generates class-balanced spatiotemporal federated features through Prototypical Gaussian Sampling, subsequently refined via a Motion-aware Differential Loss to capture kinematic properties. These features enable retraining of a globally debiased recognizer that achieves accuracy comparable to real-data-trained models. Further stabilization is achieved through Knowledge Matching, which enhances global understanding. Experiments under natural and label heterogeneity confirm that CSAR outperforms state-of-the-art methods.

Medical image segmentation demands the aggregation of global and local feature representations, posing a challenge for current methodologies in handling both long-range and short-range feature interactions. Recently, vision mamba (ViM) models have emerged as promising solutions for addressing model complexities by excelling in long-range feature iterations with linear complexity. However, existing ViM approaches overlook the importance of preserving short-range local dependencies by directly flattening spatial tokens and are constrained by fixed scanning patterns that limit the capture of dynamic spatial context information. To address these challenges, we introduce a simple yet effective method named context clustering ViM (CCViM), which incorporates a context clustering module within the existing ViM models to segment image tokens into distinct windows for adaptable local clustering. Our method effectively combines long-range and short-range feature interactions, thereby enhancing spatial contextual representations for medical image segmentation tasks. Extensive experimental evaluations on diverse public datasets, i.e., Kumar, CPM17, ISIC17, ISIC18, and Synapse, demonstrate the superior performance of our method compared to current state-of-the-art methods. Our code can be found at https://github.com/zymissy/CCViM.

Zero-Shot Skeleton-Based Action Recognition (ZSSAR) is an emerging research field focused on developing alignment models that connect skeleton movements with action definitions, thus enabling generalization to unobserved actions. Current methods often employ generative models to reconstruct cross-modal features or enhance mutual information across modalities for alignment. However, when applied to unseen action categories, these models often neglect the inherent consistency among basic actions, thereby diminishing their generalization capabilities. Furthermore, imprecise annotations fail to capture the rich semantic details of actions, resulting in misalignment. Inspired by human cognitive processes and chain of thought, we argue that integrating prior information about human actions with intrinsic commonality knowledge of basic actions is essential for ZSSAR. To actualize this, we propose a novel method termed Prompt-guided Prototype-aware Commonality and Discrimination Learning (PP-CDL). This method utilize the comprehensive world knowledge contained in LLMs, employing tailored prompts to partition seen action categories into distinct, non-overlapping prototype spaces that embody the commonality knowledge of basic actions. Subsequently, we introduce the Inter- and Intra-Prototype Discriminating (I2PD) module and the Intra-Prototype Commonality Mining (IPCM) module. The I2PD amplifies the distinctiveness of knowledge within prototypes, furnishing a personalized search space for the recognition of unseen actions. In contrast, the IPCM models the shared commonality concept within prototypes, bolstering the consistency between skeleton action representations and corresponding text knowledge representations. Experiments on different skeleton action benchmarks demonstrate the significant improvement of our method over existing alternatives.

This article proposes a new hashing framework named relational consistency induced self-supervised hashing (RCSH) for large-scale image retrieval. To capture the potential semantic structure of data, RCSH explores the relational consistency between data samples in different spaces, which learns reliable data relationships in the latent feature space and then preserves the learned relationships in the Hamming space. The data relationships are uncovered by learning a set of prototypes that group similar data samples in the latent feature space. By uncovering the semantic structure of the data, meaningful data-to-prototype and data-to-data relationships are jointly constructed. The data-to-prototype relationships are captured by constraining the prototype assignments generated from different augmented views of an image to be the same. Meanwhile, these data-to-prototype relationships are preserved to learn informative compact hash codes by matching them with these reliable prototypes. To accomplish this, a novel dual prototype contrastive loss is proposed to maximize the agreement of prototype assignments in the latent feature space and Hamming space. The data-to-data relationships are captured by enforcing the distribution of pairwise similarities in the latent feature space and Hamming space to be consistent, which makes the learned hash codes preserve meaningful similarity relationships. Extensive experimental results on four widely used image retrieval datasets demonstrate that the proposed method significantly outperforms the state-of-the-art methods. Besides, the proposed method achieves promising performance in out-of-domain retrieval tasks, which shows its good generalization ability. The source code and models are available at https://github.com/IMAG-LuJin/RCSH.

Deep learning-based models have been shown to outperform human beings in many computer vision tasks with massive available labeled training data in learning. However, humans have an amazing ability to easily recognize images of novel categories by browsing only a few examples of these categories. In this case, few-shot learning comes into being to make machines learn from extremely limited labeled examples. One possible reason why human beings can well learn novel concepts quickly and efficiently is that they have sufficient visual and semantic prior knowledge. Toward this end, this work proposes a novel knowledge-guided semantic transfer network (KSTNet) for few-shot image recognition from a supplementary perspective by introducing auxiliary prior knowledge. The proposed network jointly incorporates vision inferring, knowledge transferring, and classifier learning into one unified framework for optimal compatibility. A category-guided visual learning module is developed in which a visual classifier is learned based on the feature extractor along with the cosine similarity and contrastive loss optimization. To fully explore prior knowledge of category correlations, a knowledge transfer network is then developed to propagate knowledge information among all categories to learn the semantic-visual mapping, thus inferring a knowledge-based classifier for novel categories from base categories. Finally, we design an adaptive fusion scheme to infer the desired classifiers by effectively integrating the above knowledge and visual information. Extensive experiments are conducted on two widely used Mini-ImageNet and Tiered-ImageNet benchmarks to validate the effectiveness of KSTNet. Compared with the state of the art, the results show that the proposed method achieves favorable performance with minimal bells and whistles, especially in the case of one-shot learning.

Knowledge distillation-based anomaly detection (KDAD) methods rely on the teacher-student paradigm to detect and segment anomalous regions by contrasting the unique features extracted by both networks. However, existing KDAD methods suffer from two main limitations: 1) the student network can effortlessly replicate the teacher network's representations and 2) the features of the teacher network serve solely as a "reference standard" and are not fully leveraged. Toward this end, we depart from the established paradigm and instead propose an innovative approach called asymmetric distillation postsegmentation (ADPS). Our ADPS employs an asymmetric distillation paradigm that takes distinct forms of the same image as the input of the teacher-student networks, driving the student network to learn discriminating representations for anomalous regions. Meanwhile, a customized Weight Mask Block (WMB) is proposed to generate a coarse anomaly localization mask that transfers the distilled knowledge acquired from the asymmetric paradigm to the teacher network. Equipped with WMB, the proposed postsegmentation module (PSM) can effectively detect and segment abnormal regions with fine structures and clear boundaries. Experimental results demonstrate that the proposed ADPS outperforms the state-of-the-art methods in detecting and segmenting anomalies. Surprisingly, ADPS significantly improves average precision (AP) metric by $\mathbf {9}\%$ and $\mathbf {20}\%$ on the MVTec anomaly detection (AD) and KolektorSDD2 datasets, respectively.

Part feature learning plays a crucial role in achieving fine-grained semantic understanding in unsupervised vehicle re-identification. However, existing approaches directly model part and global features, which can easily lead to severe gradient vanishing issues due to their unequal feature information and unreliable pseudo-labels. To address this problem, in this article, we propose a triplet contrastive representation learning (TCRL) framework, which leverages cluster features to bridge the part features and global features for unsupervised vehicle re-identification. Specifically, TCRL devises three memory banks to store the instance/cluster features and proposes a proxy contrastive loss (PCL) to make contrastive learning between adjacent memory banks, thus presenting the associations between the part and global features as a transition of the part-cluster and cluster-global associations. Since the cluster memory bank copes with all the vehicle features, it can summarize them into a discriminative feature representation. To deeply exploit the instance/cluster information, TCRL proposes two additional loss functions. For the instance-level feature, a hybrid contrastive loss (HCL) re-defines the sample correlations by approaching the positive instance features and pushing all negative instance features away. For the cluster-level feature, a weighted regularization cluster contrastive loss (WRCCL) refines the pseudo labels by penalizing the mislabeled images according to the instance similarity. Extensive experiments show that TCRL outperforms many state-of-the-art unsupervised vehicle re-identification approaches.

Learning correspondences aims to find correct correspondences (inliers) from the initial correspondence set with an uneven correspondence distribution and a low inlier rate, which can be regarded as graph data. Recent advances usually use graph neural networks (GNNs) to build a single type of graph or simply stack local graphs into the global one to complete the task. But they ignore the complementary relationship between different types of graphs, which can effectively capture potential relationships among sparse correspondences. To address this problem, we propose MGNet to effectively combine multiple complementary graphs. To obtain information integrating implicit and explicit local graphs, we construct local graphs from implicit and explicit aspects and combine them effectively, which is used to build a global graph. Moreover, we propose Graph Soft Degree Attention (GSDA) to make full use of all sparse correspondence information at once in the global graph, which can capture and amplify discriminative features. Extensive experiments demonstrate that MGNet outperforms state-of-the-art methods in different visual tasks. The code is provided in https://github.com/DAILUANYUAN/MGNet-2024AAAI.

Semantic scene completion (SSC) aims to predict complete 3D voxel occupancy and semantics from a single-view RGB-D image, and recent SSC methods commonly adopt multi-modal inputs. However, our investigation reveals two limitations: ineffective feature learning from single modalities and overfitting to limited datasets. To address these issues, this paper proposes a novel SSC framework - Adversarial Modality Modulation Network (AMMNet) - with a fresh perspective of optimizing gradient updates. The proposed AMMNet introduces two core modules: a cross-modal modulation enabling the interdependence of gradient flows between modalities, and a customized adversarial training scheme leveraging dynamic gradient competition. Specifically, the cross-modal modulation adaptively re-calibrates the features to better excite representation potentials from each single modality. The adversarial training employs a minimax game of evolving gradients, with customized guidance to strengthen the generator's perception of visual fidelity from both geometric completeness and semantic correctness. Extensive experimental results demonstrate that AMMNet outperforms state-of-the-art SSC methods by a large margin, providing a promising direction for improving the effectiveness and generalization of SSC methods. Our code is available at this link.

Two-view correspondence pruning aims to accurately remove incorrect correspondences (outliers) from initial ones. Graph Neural Networks (GNNs) incorporated by Multilayer Perceptrons (MLPs) are treated as a powerful manner to handle sparse and unevenly distributed data. However, the expression capability of correspondence features obtained by MLPs is limited by their inherent insufficient of context information. In addition, previous works directly utilize the outputs of off-the-shelf GNNs, thus leading to confusion between sparse correspondence attribute features and their global structural information. To alleviate these issues, we propose a two-view correspondence pruning network TrGa. Specifically, we firstly use complete Transformer structures instead of context-agnostic MLPs to capture correspondence features with global context information and stronger expression capability. After that, we introduce the Concatenation Graph Node and Global Structure (CGNS) block to separately capture the interaction patterns among sparse correspondence attribute features and the global structural information among them, which can prevent their confusion. Finally, the proposed Feature Dimension Transformation and Enhancement (FDTE) block is applied for dimension transformation and feature augmentation. Additionally, we propose an efficient variant C-TrGa, in which the similarity matrix of the proposed C-Transformer is computed along the channel dimension. Extensive experiments demonstrate that the proposed TrGa and C-TrGa outperform state-of-the-art methods in different computer vision tasks.

Fine-grained image retrieval (FGIR) is to learn visual representations that distinguish visually similar objects while maintaining generalization. Existing methods propose to generate discriminative features, but rarely consider the particularity of the FGIR task itself. This paper presents a meticulous analysis leading to the proposal of practical guidelines to identify subcategory-specific discrepancies and generate discriminative features to design effective FGIR models. These guidelines include emphasizing the object (G1), highlighting subcategory-specific discrepancies (G2), and employing effective training strategy (G3). Following G1 and G2, we design a novel Dual Visual Filtering mechanism for the plain visual transformer, denoted as DVF, to capture subcategory-specific discrepancies. Specifically, the dual visual filtering mechanism comprises an object-oriented module and a semantic-oriented module. These components serve to magnify objects and identify discriminative regions, respectively. Following G3, we implement a discriminative model training strategy to improve the discriminability and generalization ability of DVF. Extensive analysis and ablation studies confirm the efficacy of our proposed guidelines. Without bells and whistles, the proposed DVF achieves state-of-the-art performance on three widely-used fine-grained datasets in closed-set and open-set settings.

Video frame interpolation is a critical component of video streaming, a vibrant research area dealing with requests of both service providers and users. However, existing methods cannot handle changing video resolutions while improving user perceptual quality. We aim to unleash the multifaceted knowledge yielded by the hierarchical views at multiple scales in a pyramid network. Specifically, we build a dual-view pyramid network by introducing pyramidal dual-view correspondence matching. It compels each scale to actively seek knowledge in view of both the current scale and a coarser scale, conducting robust correspondence matching by considering neighboring scales. Meanwhile, an auxiliary multi-scale collaborative supervision is devised to enforce the exchange of knowledge among scales and thus reduce error propagation from coarse to fine scales. Based on the robust capture of video dynamics via pyramidal dual-view correspondence matching, we further construct a pyramidal refinement module that formulates frame refinement as progressive latent representation generations by developing flow-guided cross-scale attention for feature fusion among frames. The proposed method is able to improve the perceptual quality on several benchmarks of varying video resolutions, while keeping low distortion and a compact model size.

In this paper, we study the problem of Text-to-Image Person Re-identification (TIReID), which aims to find images of the same identity described by a text sentence from a pool of candidate images. Benefiting from Vision-Language Pre-training, such as CLIP (Contrastive Language-Image Pretraining), the TIReID techniques have achieved remarkable progress recently. However, most existing methods only focus on instance-level matching and ignore identity-level matching, which involves associating multiple images and texts belonging to the same person. In this paper, we propose a novel prototypical prompting framework (Propot) designed to simultaneously model instance-level and identity-level matching for TIReID. Our Propot transforms the identity-level matching problem into a prototype learning problem, aiming to learn identity-enriched prototypes. Specifically, Propot works by 'initialize, adapt, enrich, then aggregate'. We first use CLIP to generate high-quality initial prototypes. Then, we propose a domain-conditional prototypical prompting (DPP) module to adapt the prototypes to the TIReID task using task-related information. Further, we propose an instance-conditional prototypical prompting (IPP) module to update prototypes conditioned on intra-modal and inter-modal instances to ensure prototype diversity. Finally, we design an adaptive prototype aggregation module to aggregate these prototypes, generating final identity-enriched prototypes. With identity-enriched prototypes, we diffuse its rich identity information to instances through prototype-to-instance contrastive loss to facilitate identity-level matching. Extensive experiments conducted on three benchmarks demonstrate the superiority of Propot compared to existing TIReID methods.

Diffusion models represent a promising avenue for image generation, having demonstrated competitive performance in pose-guided person image generation. However, existing methods are limited to generating target images from a source image and a target pose, overlooking two critical user scenarios: generating multiple target images with different poses simultaneously and generating target images from multi-view source images. To overcome these limitations, we propose IMAGPose, a unified conditional framework for pose-guided image generation, which incorporates three pivotal modules: a feature-level conditioning (FLC) module, an image-level conditioning (ILC) module, and a cross-view attention (CVA) module. Firstly, the FLC module combines the low-level texture feature from the VAE encoder with the high-level semantic feature from the image encoder, addressing the issue of missing detail information due to the absence of a dedicated person image feature extractor. Then, the ILC module achieves an alignment of images and poses to adapt to flexible and diverse user scenarios by injecting a variable number of source image conditions and introducing a masking strategy. Finally, the CVA module introduces decomposing global and local cross-attention, ensuring local fidelity and global consistency of the person image when multiple source image prompts. The three modules of IMAGPose work together to unify the task of person image generation under various user scenarios. Extensive experiment results demonstrate the consistency and photorealism of our proposed IMAGPose under challenging user scenarios. The code and model will be available at https://github.com/muzishen/IMAGPose.

Federated Instruction Tuning (FIT) advances collaborative training on decentralized data, crucially enhancing model's capability and safeguarding data privacy. However, existing FIT methods are dedicated to handling data heterogeneity across different clients (i.e., client-aware data heterogeneity), while ignoring the variation between data from different domains (i.e., domain-aware data heterogeneity). When scarce data needs supplementation from related fields, these methods lack the ability to handle domain heterogeneity in cross-domain training. This leads to domain-information catastrophic forgetting in collaborative training and therefore makes model perform sub-optimally on the individual domain. To address this issue, we introduce DoFIT, a new Domain-aware FIT framework that alleviates catastrophic forgetting through two new designs. First, to reduce interference information from the other domain, DoFIT finely aggregates overlapping weights across domains on the inter-domain server side. Second, to retain more domain information, DoFIT initializes intra-domain weights by incorporating inter-domain information into a less-conflicted parameter space. Experimental results on diverse datasets consistently demonstrate that DoFIT excels in cross-domain collaborative training and exhibits significant advantages over conventional FIT methods in alleviating catastrophic forgetting. Code is available at [this link](https://github.com/1xbq1/DoFIT).

Visual grounding (VG) aims to locate a specific target in an image based on a given language query. The discriminative information from context is important for distinguishing the target from other objects, particularly for the targets that have the same category as others. However, most previous methods underestimate such information. Moreover, they are usually designed for the standard scene (without any novel object), which limits their generalization to the open-vocabulary scene. In this paper, we propose a novel framework with context disentangling and prototype inheriting for robust visual grounding to handle both scenes. Specifically, the context disentangling disentangles the referent and context features, which achieves better discrimination between them. The prototype inheriting inherits the prototypes discovered from the disentangled visual features by a prototype bank to fully utilize the seen data, especially for the open-vocabulary scene. The fused features, obtained by leveraging Hadamard product on disentangled linguistic and visual features of prototypes to avoid sharp adjusting the importance between the two types of features, are then attached with a special token and feed to a vision Transformer encoder for bounding box regression. Extensive experiments are conducted on both standard and open-vocabulary scenes. The performance comparisons indicate that our method outperforms the state-of-the-art methods in both scenarios.

Stereo matching is a fundamental building block for many vision and robotics applications. An informative and concise cost volume representation is vital for stereo matching of high accuracy and efficiency. In this article, we present a novel cost volume construction method, named attention concatenation volume (ACV), which generates attention weights from correlation clues to suppress redundant information and enhance matching-related information in the concatenation volume. The ACV can be seamlessly embedded into most stereo matching networks, the resulting networks can use a more lightweight aggregation network and meanwhile achieve higher accuracy. We further design a fast version of ACV to enable real-time performance, named Fast-ACV, which generates high likelihood disparity hypotheses and the corresponding attention weights from low-resolution correlation clues to significantly reduce computational and memory cost and meanwhile maintain a satisfactory accuracy. Furthermore, we design a highly accurate network ACVNet and a real-time network Fast-ACVNet based on our ACV and Fast-ACV respectively, which achieve state-of-the-art performance on several benchmarks.

How to effectively explore the colors of exemplars and propagate them to colorize each frame is vital for exemplar-based video colorization. In this article, we present a BiSTNet to explore colors of exemplars and utilize them to help video colorization by a bidirectional temporal feature fusion with the guidance of semantic image prior. We first establish the semantic correspondence between each frame and the exemplars in deep feature space to explore color information from exemplars. Then, we develop a simple yet effective bidirectional temporal feature fusion module to propagate the colors of exemplars into each frame and avoid inaccurate alignment. We note that there usually exist color-bleeding artifacts around the boundaries of important objects in videos. To overcome this problem, we develop a mixed expert block to extract semantic information for modeling the object boundaries of frames so that the semantic image prior can better guide the colorization process. In addition, we develop a multi-scale refinement block to progressively colorize frames in a coarse-to-fine manner. Extensive experimental results demonstrate that the proposed BiSTNet performs favorably against state-of-the-art methods on the benchmark datasets and real-world scenes. Moreover, the BiSTNet obtains one champion in NTIRE 2023 video colorization challenge (Kang et al. 2023).

Occlusion perturbation presents a significant challenge in person re-identification (re-ID), and existing methods that rely on external visual cues require additional computational resources and only consider the issue of missing information caused by occlusion. In this paper, we propose a simple yet effective framework, termed Erasing, Transforming, and Noising Defense Network (ETNDNet), which treats occlusion as a noise disturbance and solves occluded person re-ID from the perspective of adversarial defense. In the proposed ETNDNet, we introduce three strategies: Firstly, we randomly erase the feature map to create an adversarial representation with incomplete information, enabling adversarial learning of identity loss to protect the re-ID system from the disturbance of missing information. Secondly, we introduce random transformations to simulate the position misalignment caused by occlusion, training the extractor and classifier adversarially to learn robust representations immune to misaligned information. Thirdly, we perturb the feature map with random values to address noisy information introduced by obstacles and non-target pedestrians, and employ adversarial gaming in the re-ID system to enhance its resistance to occlusion noise. Without bells and whistles, ETNDNet has three key highlights: (i) it does not require any external modules with parameters, (ii) it effectively handles various issues caused by occlusion from obstacles and non-target pedestrians, and (iii) it designs the first GAN-based adversarial defense paradigm for occluded person re-ID. Extensive experiments on six public datasets fully demonstrate the effectiveness, superiority, and practicality of the proposed ETNDNet. The code will be released at https://github.com/nengdong96/ETNDNet.

The recovery of 3D human mesh from monocular images has significantly been developed in recent years. However, existing models usually ignore spatial and temporal information, which might lead to mesh and image misalignment and temporal discontinuity. For this reason, we propose a novel Spatio-Temporal Alignment Fusion (STAF) model. As a video-based model, it leverages coherence clues from human motion by an attention-based Temporal Coherence Fusion Module (TCFM). As for spatial mesh-alignment evidence, we extract fine-grained local information through predicted mesh projection on the feature maps. Based on the spatial features, we further introduce a multi-stage adjacent Spatial Alignment Fusion Module (SAFM) to enhance the feature representation of the target frame. In addition to the above, we propose an Average Pooling Module (APM) to allow the model to focus on the entire input sequence rather than just the target frame. This method can remarkably improve the smoothness of recovery results from video. Extensive experiments on 3DPW, MPII3D, and H36M demonstrate the superiority of STAF. We achieve a state-of-the-art trade-off between precision and smoothness. Our code and more video results are on the project pagehttps://yw0208.github.io/staf/.

Despite recent advancements in masked skeleton modeling and visual-language pre-training, no method has yet been proposed to explore capturing and utilizing the rich semantic information embedded in both modalities for enhanced action recognition. To address this challenge, we propose a novel Motion-Aware Mask Feature Reconstruction (MMFR) method for the challenging task of skeleton-based action recognition. MMFR ingeniously integrates masked skeleton feature reconstruction with visual-language pre-trained model within a consolidated framework, aiming to leverage the synergistic potential of both domains. Specifically, It employs visual-language model to infuse semantic understanding into the skeleton feature reconstruction process via probability distribution distillation. Moreover, we introduce a multi-granularity semantic contrast module that refines vision-text alignment precision and augments contextual information for accurate mask reconstruction. Extensive experiments demonstrate MMFR’s superiority in skeleton-based action recognition, as well as its efficacy in zero-shot scenarios.

The image-level label has prevailed in weakly supervised semantic segmentation tasks due to its easy availability. Since image-level labels can only indicate the existence or absence of specific categories of objects, visualization-based techniques have been widely adopted to provide object location clues. Considering class activation maps (CAMs) can only locate the most discriminative part of objects, recent approaches usually adopt an expansion strategy to enlarge the activation area for more integral object localization. However, without proper constraints, the expanded activation will easily intrude into the background region. In this paper, we propose spatial structure constraints (SSC) for weakly supervised semantic segmentation to alleviate the unwanted object over-activation of attention expansion. Specifically, we propose a CAM-driven reconstruction module to directly reconstruct the input image from deep CAM features, which constrains the diffusion of last-layer object attention by preserving the coarse spatial structure of the image content. Moreover, we propose an activation self-modulation module to refine CAMs with finer spatial structure details by enhancing regional consistency. Without external saliency models to provide background clues, our approach achieves 72.7% and 47.0% mIoU on the PASCAL VOC 2012 and COCO datasets, respectively, demonstrating the superiority of our proposed approach. The source codes and models have been made available at https://github.com/NUST-Machine-Intelligence-Laboratory/SSC.

Recognizing actions performed on unseen objects, known as Compositional Action Recognition (CAR), has attracted increasing attention in recent years. The main challenge is to overcome the distribution shift of "action-objects" pairs between the training and testing sets. Previous works for CAR usually introduce extra information (e.g. bounding box) to enhance the dynamic cues of video features. However, these approaches do not essentially eliminate the inherent inductive bias in the video, which can be regarded as the stumbling block for model generalization. Because the video features are usually extracted from the visually cluttered areas in which many objects cannot be removed or masked explicitly. To this end, this work attempts to implicitly accomplish semantic-level decoupling of "object-action" in the high-level feature space. Specifically, we propose a novel Semantic-Decoupling Transformer framework, dubbed as DeFormer, which contains two insightful sub-modules: Objects-Motion Decoupler (OMD) and Semantic-Decoupling Constrainer (SDC). In OMD, we initialize several learnable tokens incorporating annotation priors to learn an instance-level representation and then decouple it into the appearance feature and motion feature in high-level visual space. In SDC, we use textual information in the high-level language space to construct a dual-contrastive association to constrain the decoupled appearance feature and motion feature obtained in OMD. Extensive experiments verify the generalization ability of DeFormer. Specifically, compared to the baseline method, DeFormer achieves absolute improvements of 3%, 3.3%, and 5.4% under three different settings on STH-ELSE, while corresponding improvements on EPIC-KITCHENS-55 are 4.7%, 9.2%, and 4.4%. Besides, DeFormer gains state-of-the-art results either on ground-truth or detected annotations.

Fine-grained visual classification aims to classify similar sub-categories with the challenges of large variations within the same sub-category and high visual similarities between different sub-categories. Recently, methods that extract semantic parts of the discriminative regions have attracted increasing attention. However, most existing methods extract the part features via rectangular bounding boxes by object detection module or attention mechanism, which makes it difficult to capture the rich shape information of objects. In this paper, we propose a novel Multi-Granularity Part Sampling Attention (MPSA) network for fine-grained visual classification. First, a novel multi-granularity part retrospect block is designed to extract the part information of different scales and enhance the high-level feature representation with discriminative part features of different granularities. Then, to extract part features of various shapes at each granularity, we propose part sampling attention, which can sample the implicit semantic parts on the feature maps comprehensively. The proposed part sampling attention not only considers the importance of sampled parts but also adopts the part dropout to reduce the overfitting issue. In addition, we propose a novel multi-granularity fusion method to highlight the foreground features and suppress the background noises with the assistance of the gradient class activation map. Experimental results demonstrate that the proposed MPSA achieves state-of-the-art performance on four commonly used fine-grained visual classification benchmarks. The source code is publicly available at https://github.com/mobulan/MPSA.

Vision-language retrieval aims to search for similar instances in one modality based on queries from another modality. The primary objective is to learn cross-modal matching representations in a latent common space. Actually, the assumption underlying cross-modal matching is modal balance, where each modality contains sufficient information to represent the others. However, noise interference and modality insufficiency often lead to modal imbalance, making it a common phenomenon in practice. The impact of imbalance on retrieval performance remains an open question. In this paper, we first demonstrate that ultimate cross-modal matching is generally sub-optimal for cross-modal retrieval when imbalanced modalities exist. The structure of instances in the common space is inherently influenced when facing imbalanced modalities, posing a challenge to cross-modal similarity measurement. To address this issue, we emphasize the importance of meaningful structure-preserved matching. Accordingly, we propose a simple yet effective method to rebalance cross-modal matching by learning structure-preserved matching representations. Specifically, we design a novel multi-granularity cross-modal matching that incorporates structure-aware distillation alongside the cross-modal matching loss. While the cross-modal matching loss constraints instance-level matching, the structure-aware distillation further regularizes the geometric consistency between learned matching representations and intra-modal representations through the developed relational matching. Extensive experiments on different datasets affirm the superior cross-modal retrieval performance of our approach, simultaneously enhancing single-modal retrieval capabilities compared to the baseline models.

Supervised deep hashing aims to learn hash functions using label information. Existing methods learn hash functions by employing either pairwise/triplet loss to explore the point-to-point relation or center loss to explore the point-to-class relation. However, these methods overlook the collaboration between the above two kinds of relations and the hardness of pairs. In this work, we propose a novel Self-Paced Relational Contrastive Hashing (SPRCH) method with a single learning objective to capture valuable discriminative information from hard pairs using both the point-to-point and point-to-class relations. To exploit the above two kinds of relations, the Relational Contrastive Hash (RCH) loss is proposed, which ensures that each data anchor is closer to all similar data points and corresponding class centers in the Hamming space compared to dissimilar ones. Moreover, the proposed RCH loss reduces the drastic imbalance between point-to-point pairs and point-to-class pairs by rebalancing their weights. To prioritize hard pairs, a self-paced learning schedule is proposed, assigning higher weights to these pairs in the RCH loss. The self-paced learning schedule assigns dynamic weights to pairs according to their similarities and the training process. In this way, deep hash model can initially learn universal patterns from the entire set of pairs and then gradually acquire more valuable discriminative information from hard pairs. Experimental results on four widely-used image retrieval datasets demonstrate that our proposed SPRCH method significantly outperforms the state-of-the-art supervised deep hash methods.

Exploring sample relationships within each mini-batch has shown great potential for learning image representations. Existing works generally adopt the regular Transformer to model the visual content relationships, ignoring the cues of semantic/label correlations between samples. Also, they generally adopt the ‘full’ self-attention mechanism which are obviously redundant and also sensitive to the noisy samples. To overcome these issues, in this paper, we design a simple yet flexible Batch-Graph Transformer (BGFormer) for mini-batch sample representations by deeply capturing the relationships of image samples from both visual and semantic perspectives. BGFormer has three main aspects. (1) It employs a flexible graph model, termedBatch Graphto jointly encode both visual and semantic relationships of samples within each mini-batch. (2) It explores the neighborhood relationships of samples by borrowing the idea of sparse graph representation which thus performs robustly, w.r.t., noisy samples. (3) It devises a novel specific Transformer architecture that mainly adoptsdualstructure-constrained self-attention (SSA), together with graph normalization, FFN, etc, to carefully exploit the batch graph information for sample tokens (nodes) representations. As an application, we apply BGFormer to the metric learning tasks. Extensive experiments on four popular datasets demonstrate the effectiveness of the proposed model.

Large pre-trained vision-language models, such as CLIP [1], have demonstrated remarkable performance in few shot image classification. To facilitate the rapid adaptation of CLIP in downstream tasks with limited visual samples, two primary frameworks have been proposed. The first framework centers on the image encoder and introduces a trainable visual classifier after the backbone to generate logits for each object class. Nevertheless, this framework heavily depends on limited visual features extracted by the pre-trained visual encoder, which can result in over-fitting issues. The second framework aims to optimize the text encoder by using trainable soft language prompts and computing logits for each class based on the similarity between image features and optimized prompt features. However, this framework encounters the issue of imperfect alignment between the representations extracted by the image and text encoders, making it difficult to fine-tune the language prompts using visual samples. This paper proposes a Multi- Modal Prototype Regularization (MMPR) method for CLIP based few-shot fine-tuning for image classification. MMPR can address the challenges of effectively utilizing both image and text features. MMPR fine-tunes a classifier and regularizes its weights using both image-based (ImgPR) and text-based (TexPR) prototypes. ImgPR represents the mean of image representations within the same class, derived from the image encoder, to distill specific visual distribution knowledge for classifier adaptation. TexPR represents the hand-crafted prompt associated with the class, derived from the text encoder, to incorporate general encyclopedic knowledge and mitigate visual over-fitting. MMPR significantly leverages both image and text information without increasing computational complexity during the inference stage compared to existing methods. Experimental results on various challenging public benchmarks demonstrate the superiority of the proposed MMPR method over state-of-the-art methods.

Hashing-based fine-grained image retrieval pursues learning diverse local features to generate inter-class discriminative hash codes. However, existing fine-grained hash methods with attention mechanisms usually tend to just focus on a few obvious areas, which misguides the network to over-fit some salient features. Such a problem raises two main limitations. 1) It overlooks some subtle local features, degrading the generalization capability of learned embedding. 2) It causes the over-activation of some hash bits correlated to salient features, which breaks the binary code balance and further weakens the discrimination abilities of hash codes. To address these limitations of the over-fitting problem, we propose a novel hash framework fromCausalFeature learning toBinary-injectedHash learning (CFBH), which captures various local information and suppresses over-activated hash bits simultaneously. For causal feature learning, we adopt causal inference theory to alleviate the bias towards the salient regions in fine-grained images. In detail, we obtain local features from the feature map and combine this local information with original image information followed by this theory. Theoretically, these fused embeddings help the network to re-weight the retrieval effort of each local feature and exploit more subtle variations without observational bias. For binary-injected hash learning, we propose a Binary Noise Injection (BNI) module inspired by Dropout. The BNI module not only mitigates over-activation to particular bits, but also makes hash codes uncorrelated and balanced in the Hamming space. Extensive experimental results on six popular fine-grained image datasets demonstrate the superiority of CFBH over several State-of-the-Art methods.

Supervised classification of hyperspectral image (HSI) is generally required to obtain better performance in spectral-spatial feature learning by fully using complex pixel- and superpixel-level interdependencies with small labeled samples. Limited by the local regular convolutions, convolutional neural networks (CNNs) can only exploit information from the short-range Euclidean neighbors of a target, hindering the effectiveness of feature representation. In contrast, graph convolutional networks (GCNs) can learn long-range dependencies between non-Euclidean neighbors but usually require the input of a full graph constructed from a whole HSI, making GCNs must be trained in a full-batch manner with tremendous computational consumption. In this work, we propose a composite neighbor-aware convolutional metric network (CNCMN), aiming to learn each target's representation from its composite neighbors (i.e., both Euclidean and non-Euclidean neighbors) in a batchwise manner. Specifically, for each target in an HSI, its Euclidean neighbors are the pixels in the local square region centered on itself, and its non-Euclidean neighbors are several related nodes selected from the constructed full graph. Correspondingly, a composite convolution (CoConv) is proposed by coupling an image convolution and a graph convolution, which can perform flexible convolutions on those composite neighbors and extract adaptively fused features from them. Besides, to further boost classification, we also propose a mini-batch metric classifier to dynamically optimize interclass and intraclass distances of samples batch by batch, which is then combined with the CoConv to form the mini-batch CNCMN. Extensive experiments on three real-world HSIs demonstrate the advantages of the proposed method over mini-batch deep learning algorithms and have obtained the state-of-the-art performance in these fields. The code is available at: https://github.com/qichaoliu/HSI-CNCMN.

Thanks to the advantages of the friendly annotations and the satisfactory performance, weakly-supervised semantic segmentation (WSSS) approaches have been extensively studied. Recently, the single-stage WSSS (SS-WSSS) was awakened to alleviate problems of the expensive computational costs and the complicated training procedures in multistage WSSS. However, the results of such an immature model suffer from problems of background incompleteness and object incompleteness. We empirically find that they are caused by the insufficiency of the global object context and the lack of local regional contents, respectively. Under these observations, we propose an SS-WSSS model with only the image-level class label supervisions, termed weakly supervised feature coupling network (WS-FCN), which can capture the multiscale context formed from the adjacent feature grids, and encode the fine-grained spatial information from the low-level features into the high-level ones. Specifically, a flexible context aggregation (FCA) module is proposed to capture the global object context in different granular spaces. Besides, a semantically consistent feature fusion (SF2) module is proposed in a bottom-up parameter-learnable fashion to aggregate the fine-grained local contents. Based on these two modules, WS-FCN lies in a self-supervised end-to-end training fashion. Extensive experimental results on the challenging PASCAL VOC 2012 and MS COCO 2014 demonstrate the effectiveness and efficiency of WS-FCN, which can achieve state-of-the-art results by 65.02% and 64.22% mIoU on PASCAL VOC 2012 val set and test set, 34.12% mIoU on MS COCO 2014 val set, respectively. The code and weight have been released at:WS-FCN.

Text-based person search (TBPS) is a challenging task that aims to search pedestrian images with the same identity from an image gallery given a query text. In recent years, TBPS has made remarkable progress, and state-of-the-art (SOTA) methods achieve superior performance by learning local fine-grained correspondence between images and texts. However, most existing methods rely on explicitly generated local parts to model fine-grained correspondence between modalities, which is unreliable due to the lack of contextual information or the potential introduction of noise. Moreover, the existing methods seldom consider the information inequality problem between modalities caused by image-specific information. To address these limitations, we propose an efficient joint multilevel alignment network (MANet) for TBPS, which can learn aligned image/text feature representations between modalities at multiple levels, and realize fast and effective person search. Specifically, we first design an image-specific information suppression (ISS) module, which suppresses image background and environmental factors by relation-guided localization (RGL) and channel attention filtration (CAF), respectively. This module effectively alleviates the information inequality problem and realizes the alignment of information volume between images and texts. Second, we propose an implicit local alignment (ILA) module to adaptively aggregate all pixel/word features of image/text to a set of modality-shared semantic topic centers and implicitly learn the local fine-grained correspondence between modalities without additional supervision and cross-modal interactions. Also, a global alignment (GA) is introduced as a supplement to the local perspective. The cooperation of global and local alignment modules enables better semantic alignment between modalities. Extensive experiments on multiple databases demonstrate the effectiveness and superiority of our MANet.

Video-Text pre-training aims at learning transferable representations from large-scale video-text pairs via aligning the semantics between visual and textual information. State-of-the-art approaches extract visual features from raw pixels in an end-to-end fashion. However, these methods operate at frame-level directly and thus overlook the spatio-temporal structure of objects in video, which yet has a strong synergy with nouns in textual descriptions. In this work, we propose a simple yet effective module for video-text representation learning, namely RegionLearner, which can take into account the structure of objects during pre-training on large-scale video-text pairs. Given a video, our module (1) first quantizes continuous visual features via clustering patch-features into the same cluster according to content similarity, then (2) generates learnable masks to aggregate fragmentary features into regions with complete semantics, and finally (3) models the spatio-temporal dependencies between different semantic regions. In contrast to using off-the-shelf object detectors, our proposed module does not require explicit supervision and is much more computationally efficient. We pre-train the proposed approach on the public WebVid2M and CC3M datasets. Extensive evaluations on four downstream video-text retrieval benchmarks clearly demonstrate the effectiveness of our RegionLearner.

How to effectively explore spatial and temporal information is important for video deblurring. In contrast to existing methods that directly align adjacent frames without discrimination, we develop a deep discriminative spatial and temporal network to facilitate the spatial and temporal feature exploration for better video deblurring. We first develop a channel-wise gated dynamic network to adaptively explore the spatial information. As adjacent frames usually contain different contents, directly stacking features of adjacent frames without discrimination may affect the latent clear frame restoration. Therefore, we develop a simple yet effective discriminative temporal feature fusion module to obtain useful temporal features for latent frame restoration. Moreover, to utilize the information from long-range frames, we develop a wavelet-based feature propagation method that takes the discriminative temporal feature fusion module as the basic unit to effectively propagate main structures from long-range frames for better video deblurring. We show that the proposed method does not require additional alignment methods and performs favorably against state-of-the-art ones on benchmark datasets in terms of accuracy and model complexity.

Semantic Scene Completion (SSC) transforms an image of single-view depth and/or RGB 2D pixels into 3D voxels, each of whose semantic labels are predicted. SSC is a well-known ill-posed problem as the prediction model has to “imagine” what is behind the visible surface, which is usually represented by Truncated Signed Distance Function (TSDF). Due to the sensory imperfection of the depth camera, most existing methods based on the noisy TSDF estimated from depth values suffer from 1) incomplete volumetric predictions and 2) confused semantic labels. To this end, we use the ground-truth 3D voxels to generate a perfect visible surface, called TSDF-CAD, and then train a “cleaner” SSC model. As the model is noise-free, it is expected to focus more on the “imagination” of unseen voxels. Then, we propose to distill the intermediate “cleaner” knowledge into another model with noisy TSDF input. In particular, we use the 3D occupancy feature and the semantic relations of the “cleaner self” to supervise the counterparts of the “noisy self” to respectively address the above two incorrect predictions. Experimental results validate that our method improves the noisy counterparts with 3.1% IoU and 2.2% mIoU for measuring scene completion and SSC, and also achieves new state-of-the-art accuracy on the popular NYU dataset. The code is available at https://github.com/fereenwong/CleanerS.

High-performance video frame interpolation is challenging for complex scenes with diverse motion and occlusion characteristics. Existing methods, deploying off-the-shelf flow estimators to acquire initial characterizations refined by multiple subsequent models, often require heavy network architectures that are not practical for resource constrained systems. We investigate the unary potentials of the characterizations to improve efficiency. Specifically, we design a lightweight neural network to construct motion volumes via ensembles of offset approximations, and propose a spatiotemporal variance-aware loss to supervise the network learning. For network compactness, our spatiotemporal variance-supervised motion volume (SVMV) utilizes shared spatiotemporal representations via correlations among approximations, of which the diversifications are exploited to better leverage the network’s expressiveness through the spatiotemporal variances of motions and occlusions within the time interval to be interpolated. Experiments on publicly available datasets show that our method performs favorably against existing methods with a more compact network and less runtime.

We present a simple and effective Multi-scale Residual Low-Pass Filter Network (MRLPFNet) that jointly explores the image details and main structures for image deblurring. Our work is motivated by an observation that the difference between the blurry image and the clear one not only contains high-frequency contents1but also includes low-frequency information due to the influence of blur, while using the standard residual learning is less effective for modeling the main structure distorted by the blur. Considering that the low-frequency contents usually correspond to main global structures that are spatially variant, we first propose a learnable low-pass filter based on a self-attention mechanism to adaptively explore the global contexts for better modeling the low-frequency information. Then we embed it into a Residual Low-Pass Filter (RLPF) module, which involves an additional fully convolutional neural network with the standard residual learning to model the high-frequency information. We formulate the RLPF module into an end-to-end trainable network based on an encoder and decoder architecture and develop a wavelet-based feature fusion to fuse the multi-scale features. Experimental results show that our method performs favorably against state-of-the-art ones on commonly-used benchmarks.

We propose an effective lightweight dynamic local and global self-attention network (DLGSANet) to solve image super-resolution. Our method explores the properties of Transformers while having low computational costs. Motivated by the network designs of Transformers, we develop a simple yet effective multi-head dynamic local self-attention (MHDLSA) module to extract local features efficiently. In addition, we note that existing Transformers usually explore all similarities of the tokens between the queries and keys for the feature aggregation. However, using all the similarities does not effectively facilitate the high-resolution image reconstruction as not all the tokens from the queries are relevant to those in keys. To overcome this problem, we develop a sparse global self-attention (SparseGSA) module to select the most useful similarity values so that the most useful global features can be better utilized for image reconstruction. We develop a hybrid dynamic-Transformer block (HDTB) that integrates the MHDLSA and SparseGSA for both local and global feature exploration. To ease the network training, we formulate the HDTBs into a residual hybrid dynamic-Transformer group (RHDTG). By embedding the RHDTGs into an end-to-end trainable network, we show that the proposed method has fewer network parameters and lower computational costs while achieving competitive performance against state-of-the-art ones in terms of accuracy. More information is available at https://neonleexiang.github.io/DLGSANet/.

Although deep learning-based solutions have achieved impressive reconstruction performance in image super-resolution (SR), these models are generally large, with complex architectures, making them incompatible with low-power devices with many computational and memory constraints. To overcome these challenges, we propose a spatially-adaptive feature modulation (SAFM) mechanism for efficient SR design. In detail, the SAFM layer uses independent computations to learn multi-scale feature representations and aggregates these features for dynamic spatial modulation. As the SAFM prioritizes exploiting non-local feature dependencies, we further introduce a convolutional channel mixer (CCM) to encode local contextual information and mix channels simultaneously. Extensive experimental results show that the proposed method is 3× smaller than state-of-the-art efficient SR methods, e.g., IMDN, and yields comparable performance with much less memory usage. Our source codes and pre-trained models are available at: https://github.com/sunny2109/SAFMN.

Due to the scarcity of manually annotated data required for fine-grained video understanding, few-shot fine-grained (FS-FG) action recognition has gained significant attention, with the aim of classifying novel fine-grained action categories with only a few labeled instances. Despite the progress made in FS coarse-grained action recognition, current approaches encounter two challenges when dealing with the fine-grained action categories: the inability to capture subtle action details and the insufficiency of learning from limited data that exhibit high intra-class variance and inter-class similarity. To address these limitations, we propose M3Net, a matching-based framework for FS-FG action recognition, which incorporates multi-view encoding, multi-view matching, and multi-view fusion to facilitate embedding encoding, similarity matching, and decision making across multiple viewpoints.Multi-view encoding captures rich contextual details from the intra-frame, intra-video, and intra-episode perspectives, generating customized higher-order embeddings for fine-grained data.Multi-view matching integrates various matching functions enabling flexible relation modeling within limited samples to handle multi-scale spatio-temporal variations by leveraging the instance-specific, category-specific, and task-specific perspectives. Multi-view fusion consists of matching-predictions fusion and matching-losses fusion over the above views, where the former promotes mutual complementarity and the latter enhances embedding generalizability by employing multi-task collaborative learning. Explainable visualizations and experimental results on three challenging benchmarks demonstrate the superiority of M3Net in capturing fine-grained action details and achieving state-of-the-art performance for FS-FG action recognition.

Spatio-temporal Action Detection (SAD) aims to recognize the multi-class actions, and meanwhile locate their spatio-temporal occurrence in untrimmed videos. Besides relying on the inherent inter-actor interactions, most previous SAD approaches model actor interactions between multi-actors and the whole frames or special parts (e.g., objects/hands). However, such approaches are relatively graceless by 1) roughly treating all various actors to equivalently interact with frames/parts or by 2) sumptuously borrowing multiple costly detectors to acquire the special parts. To solve the above dilemma, we propose a novel Foreground/Background-masked Interaction Learning (dubbed as FBI Learning) framework to learn the multi-actor features by attentively interacting with the hands-down foreground and background frames. Specifically, we first design a new Mask-guided Cross Attention (MCA) mechanism that calculates the masked cross-attentions to capture the compact relations between the actors and foreground/background regions. Next, we present a new Actor-guided Feature Aggregation (AFA) scheme that integrates foreground- and background-interacted actor features with the learnable actor-based weights. Finally, we construct a long-term feature bank that associates temporal context information to facilitate action classification. Extensive experiments are conducted on commonly available UCF101-24, MultiSports, and AVA v2.1/v2.2 datasets, which illustrate the competitive performance of FBI Learning against the state-of-the-art methods.

Egocentric Action Recognition (EAR) is required to recognize both the interacting objects (noun) and the motion (verb) against cluttered backgrounds with distracting objects. For capturing interacting objects, traditional approaches heavily rely on luxury object annotations or detectors, though a few works heuristically enumerate the fixed sets of verb-constrained prototypes to roughly exclude the background. For capturing motion, the inherent variations of motion duration among egocentric videos with different lengths are almost ignored. To this end, we propose a novel Slowfast Diversity-aware Prototype learning (SDP) to effectively capture interacting objects by learning compact yet diverse prototypes, and adaptively capture motion in either long-time video or short-time video. Specifically, we present a new Part-to-Prototype (P2P) scheme to learn prototypes from raw videos covering the interacting objects by refining the semantic information from part level to prototype level. Moreover, for adaptively capturing motion, we design a new Slow-Fast Context (SFC) mechanism that explores the Up/Down augmentations for the prototype representation at the semantic level to strengthen the transient dynamic information in short-time videos and eliminate the redundant dynamic information in long-time videos, which are further fine-complemented via the slow- and fast-aware attentions. Extensive experiments demonstrate SDP outperforms state-of-the-art methods on two large-scale egocentric video benchmarks, i.e., EPIC-KITCHENS-100 and EGTEA.

Text-based person retrieval is a challenging task that aims to search pedestrian images with the same identity according to language descriptions. Current methods usually indiscriminately measure the similarity between text and image by matching global visual-textual features and matched local region-word features. However, these methods underestimate the key cue role of mismatched region-word pairs and ignore the problem of low similarity between matched region-word pairs. To alleviate these issues, we propose a novel Pedestrian-specific Bipartite-aware Similarity Learning (PBSL) framework that efficiently reveals the plausible and credible levels of contribution of pedestrian-specific mismatched and matched region-word pairs towards overall similarity. Specifically, to focus on mismatched region-word pairs, we first develop a new co-interactive attention that utilizes cross-modal information to guide the extraction of pedestrian-specific information in a single modality. We then design a negative similarity regularization mechanism to use the negative similarity score as a bias to correct the overall similarity. Additionally, to enhance the contribution of matched region-word pairs, we introduce graph networks to aggregate and propagate local information of pedestrian-specific, using overall visual-textual similarity to evaluate locally matched region-word pairs for weight refinement. Finally, extensive experiments are conducted on the CUHK-PEDES, ICFG-PEDES, and RSTPReid datasets to demonstrate the competitive performance of the proposed PBSL in the text-based person retrieval task.

Text-to-image person re-identification (TIReID) retrieves pedestrian images of the same identity based on a query text. However, existing methods typically treat it as a one-to-one image-text matching problem, only focusing on the relationship between image-text pairs within a view. The many-to-many matching between image-text pairs across views under the same identity is not taken into account, which is one of the main reasons for the poor performance of existing methods. To this end, we propose a simple yet effective framework, called LCR2S, for modeling many-to-many correspondences of the same identity by learning comprehensive representations for both modalities from a novel perspective. We construct a support set for each image (text) by using other images (texts) under the same identity and design a multi-head attentional fusion module to fuse the image (text) and its support set. The resulting enriched image and text features are aligned to train a "richer" TIReID model with many-to-many correspondences. Since the support set is unavailable during inference, we propose to distill the knowledge learned by the "richer" model into a lightweight model for inference with a single image/text as input. The lightweight model focus on semantic association and reasoning of multi-view information, which can generate a comprehensive representation containing multi-view information with only a single-view input to perform accurate text-to-image retrieval during inference. In particular, we use the intra-modal features and inter-modal semantic relations of the "richer" model to supervise the lightweight model to inherit its powerful capability. Extensive experiments demonstrate the effectiveness of LCR2S, and it also achieves new state-of-the-art performance on three popular TIReID datasets.

We present compact and effective deep convolutional neural networks (CNNs) by exploring properties of videos for video deblurring. Motivated by the non-uniform blur property that not all the pixels of the frames are blurry, we develop a CNN to integrate a temporal sharpness prior (TSP) for removing blur in videos. The TSP exploits sharp pixels from adjacent frames to facilitate the CNN for better frame restoration. Observing that the motion field is related to latent frames instead of blurry ones in the image formation model, we develop an effective cascaded training approach to solve the proposed CNN in an end-to-end manner. As videos usually contain similar contents within and across frames, we propose a non-local similarity mining approach based on a self-attention method with the propagation of global features to constrain CNNs for frame restoration. We show that exploring the domain knowledge of videos can make CNNs more compact and efficient, where the CNN with the non-local spatial-temporal similarity is 3× smaller than the state-of-the-art methods in terms of model parameters while its performance gains are at least 1 dB higher in terms of PSNRs. Extensive experimental results show that our method performs favorably against state-of-the-art approaches on benchmarks and real-world videos.

In the semi-supervised skeleton-based action recognition task, obtaining more discriminative information from both labeled and unlabeled data is a challenging problem. As the current mainstream approach, contrastive learning can learn more representations of augmented data, which can be considered as the pretext task of action recognition. However, such a method still confronts three main limitations: 1) It usually learns global-granularity features that cannot well reflect the local motion information. 2) The positive/negative pairs are usually pre-defined, some of which are ambiguous. 3) It generally measures the distance between positive/negative pairs only within the same granularity, which neglects the contrasting between the cross-granularity positive and negative pairs. Toward these limitations, we propose a novel Multi-granularity Anchor-Contrastive representation Learning (dubbed as MAC-Learning) to learn multi-granularity representations by conducting inter- and intra-granularity contrastive pretext tasks on the learnable and structural-link skeletons among three types of granularities covering local, context, and global views. To avoid the disturbance of ambiguous pairs from noisy and outlier samples, we design a more reliable Multi-granularity Anchor-Contrastive Loss (dubbed as MAC-Loss) that measures the agreement/disagreement between high-confidence soft-positive/negative pairs based on the anchor graph instead of the hard-positive/negative pairs in the conventional contrastive loss. Extensive experiments on both NTU RGB+D and Northwestern-UCLA datasets show that the proposed MAC-Learning outperforms existing competitive methods in semi-supervised skeleton-based action recognition tasks.

In order to enable the model to generalize to unseen "action-objects" (compositional action), previous methods encode multiple pieces of information (i.e., the appearance, position, and identity of visual instances) independently and concatenate them for classification. However, these methods ignore the potential supervisory role of instance information (i.e., position and identity) in the process of visual perception. To this end, we present a novel framework, namely Progressive Instance-aware Feature Learning (PIFL), to progressively extract, reason, and predict dynamic cues of moving instances from videos for compositional action recognition. Specifically, this framework extracts features from foreground instances that are likely to be relevant to human actions (Position-aware Appearance Feature Extraction in Section III-B1), performs identity-aware reasoning among instance-centric features with semantic-specific interactions (Identity-aware Feature Interaction in Section III-B2), and finally predicts instances' position from observed states to force the model into perceiving their movement (Semantic-aware Position Prediction in Section III-B3). We evaluate our approach on two compositional action recognition benchmarks, namely, Something-Else and IKEA-Assembly. Our approach achieves consistent accuracy gain beyond off-the-shelf action recognition algorithms in terms of both ground truth and detected position of instances.

Group activity recognition (GAR) is a challenging task aimed at recognizing the behavior of a group of people. It is a complex inference process in which visual cues collected from individuals are integrated into the final prediction, being aware of the interaction between them. This paper goes one step further beyond the existing approaches by designing a Hierarchical Graph-based Cross Inference Network (HiGCIN), in which three levels of information, i.e., the body-region level, person level, and group-activity level, are constructed, learned, and inferred in an end-to-end manner. Primarily, we present a generic Cross Inference Block (CIB), which is able to concurrently capture the latent spatiotemporal dependencies among body regions and persons. Based on the CIB, two modules are designed to extract and refine features for group activities at each level. Experiments on two popular benchmarks verify the effectiveness of our approach, particularly in the ability to infer with multilevel visual cues. In addition, training our approach does not require individual action labels to be provided, which greatly reduces the amount of labor required in data annotation.

Automated classification of gastrointestinal endoscope images can help reduce the workload of doctors and improve the accuracy of diagnoses. The rapidly developed vision Transformer, represented by Swin Transformer, has become an impressive technique for medical image classification. However, Swin Transformer cannot capture the long-range dependency well in complex gastrointestinal endoscopy images. As a result, it fails to represent features of some widely-spread targets in digestive tract images, such as normal-z-line and esophagitis, effectively. To solve this problem, we propose a novel vision Transformer model based on hybrid shifted windows for digestive tract image classification, which can obtain both short-range and long-range dependency concurrently. Extensive experiments demonstrate the superiority of our method to the state-of-the-art methods with a classification accuracy of 95.42% on the Kvasir v2 dataset and a classification accuracy of 86.81% on the HyperKvasir dataset.

Few-shot fine-grained recognition (FS-FGR) aims to recognize novel fine-grained categories with the help of limited available samples. Undoubtedly, this task inherits the main challenges from both few-shot learning and fine-grained recognition. First, the lack of labeled samples makes the learned model easy to overfit. Second, it also suffers from high intra-class variance and low inter-class differences in the datasets. To address this challenging task, we propose a two-stage background suppression and foreground alignment framework, which is composed of a background activation suppression (BAS) module, a foreground object alignment (FOA) module, and a local-to-local (L2L) similarity metric. Specifically, the BAS is introduced to generate a foreground mask for localization to weaken background disturbance and enhance dominative foreground objects. The FOA then reconstructs the feature map of each support sample according to its correction to the query ones, which addresses the problem of misalignment between support-query image pairs. To enable the proposed method to have the ability to capture subtle differences in confused samples, we present a novel L2L similarity metric to further measure the local similarity between a pair of aligned spatial features in the embedding space. What’s more, considering that background interference brings poor robustness, we infer the pairwise similarity of feature maps using both the raw image and the refined image. Extensive experiments conducted on multiple popular fine-grained benchmarks demonstrate that our method outperforms the existing state of the art by a large margin. The source codes are available at:https://github.com/CSer-Tang-hao/BSFA-FSFG.

Hyperspectral image (HSI) classification faces the inherent challenge of small sample learning, primarily due to the difficulty in labeling vast land covers. Meta-learning, with its ability to learn transferable meta-knowledge from existing HSIs, is seen as a promising solution. However, different HSIs usually have varying distributions manifested as differing spectral wavelengths and reflectance shifts, which is often neglected in existing methods, stalling the acquisition of transferable features. To address this issue, we introduce an attribute-driven spectral alignment (ADSA) method, which parameterizes and embeds spectral attributes (i.e., spectral wavelengths and reflectance shifts) into a domain-adaptation model, aiming to decouple the domain-specific attributes of different HSIs in an unsupervised manner. After training, the parameterized attributes of source-domain (SD) HSIs can be substituted with those of the target domain (TD), allowing the decoder of ADSA to rebuild new HSIs sharing identical spectral attributes. By this means, numerous distribution-consistent labeled samples preserving inherent spectral–spatial structures can be obtained. Then, a 3-D residual prototypical network (RPN) based on 3-D convolutions and metric learning is designed to model complex structures of HSIs, which in combination with the few-shot learning (FSL) framework can extract valuable discriminative knowledge from these auxiliary samples. Finally, by applying this learned knowledge to the TD HSI, only a small number of labeled samples are required to obtain satisfactory performance. Extensive experiments on four real-world HSIs demonstrate the effectiveness of our method, and the performance outperforms several state-of-the-art methods.

With the rapid development of Deepfake synthesis technology, our information security and personal privacy have been severely threatened in recent years. To achieve a robust Deepfake detection, researchers attempt to exploit the joint spatial-temporal information in the videos, like using recurrent networks and 3D convolutional networks. However, these spatial-temporal models remain room to improve. Another general challenge for spatial-temporal models is that people do not clearly understand what these spatial-temporal models really learn. To address these two challenges, in this paper, we propose an Interpretable Spatial-Temporal Video Transformer (ISTVT), which consists of a novel decomposed spatial-temporal self-attention and a self-subtract mechanism to capture spatial artifacts and temporal inconsistency for robust Deepfake detection. Thanks to this decomposition, we propose to interpret ISTVT by visualizing the discriminative regions for both spatial and temporal dimensions via the relevance (the pixel-wise importance on the input) propagation algorithm. We conduct extensive experiments on large-scale datasets, including FaceForensics++, FaceShifter, DeeperForensics, Celeb-DF, and DFDC datasets. Our strong performance of intra-dataset and cross-dataset Deepfake detection demonstrates the effectiveness and robustness of our method, and our visualization-based interpretability offers people insights into our model.

Weakly supervised semantic segmentation (WSSS) models relying on class activation maps (CAMs) have achieved desirable performance comparing to the non-CAMs-based counterparts. However, to guarantee WSSS task feasible, we need to generate pseudo labels by expanding the seeds from CAMs which is complex and time-consuming, thus hindering the design of efficient end-to-end (single-stage) WSSS approaches. To tackle the above dilemma, we resort to the off-the-shelf and readily accessible saliency maps for directly obtaining pseudo labels given the image-level class labels. Nevertheless, the salient regions may contain noisy labels and cannot seamlessly fit the target objects, and saliency maps can only be approximated as pseudo labels for simple images containing single-class objects. As such, the achieved segmentation model with these simple images cannot generalize well to the complex images containing multi-class objects. To this end, we propose an end-to-end multi-granularity denoising and bidirectional alignment (MDBA) model, to alleviate the noisy label and multi-class generalization issues. Specifically, we propose the online noise filtering and progressive noise detection modules to tackle image-level and pixel-level noise, respectively. Moreover, a bidirectional alignment mechanism is proposed to reduce the data distribution gap at both input and output space with simple-to-complex image synthesis and complex-to-simple adversarial learning. MDBA can reach the mIoU of 69.5% and 70.2% on validation and test sets for the PASCAL VOC 2012 dataset. The source codes and models have been made available at https://github.com/NUST-Machine-Intelligence-Laboratory/MDBA.

The visual feature pyramid has shown its superiority in both effectiveness and efficiency in a variety of applications. However, current methods overly focus on inter-layer feature interactions while disregarding the importance of intra-layer feature regulation. Despite some attempts to learn a compact intra-layer feature representation with the use of attention mechanisms or vision transformers, they overlook the crucial corner regions that are essential for dense prediction tasks. To address this problem, we propose a Centralized Feature Pyramid (CFP) network for object detection, which is based on a globally explicit centralized feature regulation. Specifically, we first propose a spatial explicit visual center scheme, where a lightweight MLP is used to capture the globally long-range dependencies, and a parallel learnable visual center mechanism is used to capture the local corner regions of the input images. Based on this, we then propose a globally centralized regulation for the commonly-used feature pyramid in a top-down fashion, where the explicit visual center information obtained from the deepest intra-layer feature is used to regulate frontal shallow features. Compared to the existing feature pyramids, CFP not only has the ability to capture the global long-range dependencies but also efficiently obtain an all-round yet discriminative feature representation. Experimental results on the challenging MS-COCO validate that our proposed CFP can achieve consistent performance gains on the state-of-the-art YOLOv5 and YOLOX object detection baselines.

Text-Image Person Re-identification (TIReID) aims to retrieve the image corresponding to the given text query from a pool of candidate images. Existing methods employ prior knowledge from single-modality pre-training to facilitate learning, but lack multi-modal correspondence information. Vision-Language Pre-training, such as CLIP (Contrastive Language-Image Pretraining), can address the limitation. However, CLIP falls short in capturing fine-grained information, thereby not fully leveraging its powerful capacity in TIReID. Besides, the popular explicit local matching paradigm for mining fine-grained information heavily relies on the quality of local parts and cross-modal inter-part interaction/guidance, leading to intra-modal information distortion and ambiguity problems. Accordingly, in this paper, we propose a CLIP-driven Fine-grained information excavation framework (CFine) to fully utilize the powerful knowledge of CLIP for TIReID. To transfer the multi-modal knowledge effectively, we conduct fine-grained information excavation to mine modality-shared discriminative details for global alignment. Specifically, we propose a multi-level global feature learning (MGF) module that fully mines the discriminative local information within each modality, thereby emphasizing identity-related discriminative clues through enhanced interaction between global image (text) and informative local patches (words). MGF generates a set of enhanced global features for later inference. Furthermore, we design cross-grained feature refinement (CFR) and fine-grained correspondence discovery (FCD) modules to establish cross-modal correspondence at both coarse and fine-grained levels (image-word, sentence-patch, word-patch), ensuring the reliability of informative local patches/words. CFR and FCD are removed during inference to optimize computational efficiency. Extensive experiments on multiple benchmarks demonstrate the superior performance of our method in TIReID.

Hashing has been widely applied to multimodal retrieval on large-scale multimedia data due to its efficiency in computation and storage. In this article, we propose a novel deep semantic multimodal hashing network (DSMHN) for scalable image-text and video-text retrieval. The proposed deep hashing framework leverages 2-D convolutional neural networks (CNN) as the backbone network to capture the spatial information for image-text retrieval, while the 3-D CNN as the backbone network to capture the spatial and temporal information for video-text retrieval. In the DSMHN, two sets of modality-specific hash functions are jointly learned by explicitly preserving both intermodality similarities and intramodality semantic labels. Specifically, with the assumption that the learned hash codes should be optimal for the classification task, two stream networks are jointly trained to learn the hash functions by embedding the semantic labels on the resultant hash codes. Moreover, a unified deep multimodal hashing framework is proposed to learn compact and high-quality hash codes by exploiting the feature representation learning, intermodality similarity-preserving learning, semantic label-preserving learning, and hash function learning with different types of loss functions simultaneously. The proposed DSMHN method is a generic and scalable deep hashing framework for both image-text and video-text retrievals, which can be flexibly integrated with different types of loss functions. We conduct extensive experiments for both single-modal- and cross-modal-retrieval tasks on four widely used multimodal-retrieval data sets. Experimental results on both image-text- and video-text-retrieval tasks demonstrate that the DSMHN significantly outperforms the state-of-the-art methods.

Label noise has been a practical challenge in deep learning due to the strong capability of deep neural networks in fitting all training data. Prior literature primarily resorts to sample selection methods for combating noisy labels. However, these approaches focus on dividing samples by order sorting or threshold selection, inevitably introducing hyperparameters (e.g., selection ratio / threshold) that are hard-to-tune and dataset-dependent. To this end, we propose a simple yet effective approach named PNP (Probabilistic Noise Prediction) to explicitly model label noise. Specifically, we simultaneously train two networks, in which one predicts the category label and the other predicts the noise type. By predicting label noise probabilistically, we identify noisy samples and adopt dedicated optimization objectives accordingly. Finally, we establish a joint loss for network update by unifying the classification loss, the auxiliary constraint loss, and the in-distribution consistency loss. Comprehensive experimental results on synthetic and realworld datasets demonstrate the superiority of our proposed method. The source code and models have been made available at https://github.com/NUST-Machine-Intelligence-Laboratory/PNP.

Multimedia recommendation forms a personalized ranking task with multimedia content representations which are mostly extracted via generic encoders. However, the generic representations introduce spurious correlations --- the meaningless correlation from the recommendation perspective. For example, suppose a user bought two dresses on the same model, this co-occurrence would produce a correlation between the model and purchases, but the correlation is spurious from the view of fashion recommendation. Existing work alleviates this issue by customizing preference-aware representations, requiring high-cost analysis and design.

Scene Graph Generation (SGG) task aims to construct a graph structure to express objects and their relationships in a scene at a holistic level. Due to the neglect of heterogeneity of feature spaces between objects and relations, coupling of feature representations becomes obvious in current SGG methods, which results in large intra-class variation and inter-class ambiguity. In order to explicitly emphasize the heterogeneity in SGG, we propose a plug-and-play Heterogeneous Learning Branch (HLB), which enhances the independent representation capability of relation features. The HLB actively obscures the interconnection between objects and relation feature spaces via gradient reversal, with the assistance of a link prediction module as information barrier and an Auto Encoder for information preservation. To validate the effectiveness of HLB, we apply HLB to typical SGG methods in which the feature spaces are either homogeneous or semi-heterogeneous, and conduct evaluation on VG-150 dataset. The experimental results demonstrate that HLB significantly improves the performance of all these methods in the common evaluation criteria for SGG task.

Compositional action recognition which aims to identify the unseen combinations of actions and objects has recently attracted wide attention. Conventional methods bring in additional cues (e.g., dynamic motions of objects) to alleviate the inductive bias between the visual appearance of objects and the human action-level labels. Besides, compared with non-compositional settings, previous methods only pursue higher performance in compositional settings, which can not prove their generalization ability. To this end, we firstly rethink the problem and design a more generalized metric (namely Drop Ratio) and a more practical setting to evaluate the compositional generalization of existing action recognition algorithms. Beyond that, we propose a simple yet effective framework, Look Less Think More (LLTM), to reduce the strong association between visual objects and action-level labels (Look Less), and then discover the commonsense relationships between object categories and human actions (Think More). We test the rationality of the proposed Drop Ratio and Practical setting by comparing several popular action recognition methods on SSV2. Besides, the proposed LLTM achieves state-of-the-art performance on SSV2 with different settings.

Capturing the long-range dependencies has empirically proven to be effective on a wide range of computer vision tasks. The progressive advances on this topic have been made through the employment of the transformer framework with the help of the multi-head attention mechanism. However, the attention-based image patch interaction potentially suffers from problems of redundant interactions of intra-class patches and unoriented interactions of inter-class patches. In this paper, we propose a novel Graph Reasoning Transformer (GReaT) for image parsing to enable image patches to interact following a relation reasoning pattern. Specifically, the linearly embedded image patches are first projected into the graph space, where each node represents the implicit visual center for a cluster of image patches and each edge reflects the relation weight between two adjacent nodes. After that, global relation reasoning is performed on this graph accordingly. Finally, all nodes including the relation information are mapped back into the original space for subsequent processes. Compared to the conventional transformer, GReaT has higher interaction efficiency and a more purposeful interaction pattern. Experiments are carried out on the challenging Cityscapes and ADE20K datasets. Results show that GReaT achieves consistent performance gains with slight computational overheads on the state-of-the-art transformer baselines.

Lightweight and efficiency are critical drivers for the practical application of image super-resolution (SR) algorithms. We propose a simple and effective approach, ShuffleMixer, for lightweight image super-resolution that explores large convolution and channel split-shuffle operation. In contrast to previous SR models that simply stack multiple small kernel convolutions or complex operators to learn representations, we explore a large kernel ConvNet for mobile-friendly SR design. Specifically, we develop a large depth-wise convolution and two projection layers based on channel splitting and shuffling as the basic component to mix features efficiently. Since the contexts of natural images are strongly locally correlated, using large depth-wise convolutions only is insufficient to reconstruct fine details. To overcome this problem while maintaining the efficiency of the proposed module, we introduce Fused-MBConvs into the proposed network to model the local connectivity of different features. Experimental results demonstrate that the proposed ShuffleMixer is about $3 \times$ smaller than the state-of-the-art efficient SR methods, e.g. CARN, in terms of model parameters and FLOPs while achieving competitive performance.

Automated diagnosis of digestive tract diseases from gastrointestinal endoscopy images is of high importance for improving the diagnosis accuracy and efficiency. The current mainstream methods for image classification of digestive tract endoscopy images are based on Convolutional Neural Networks (CNNs). However, due to their inherent defects, CNNs are not strong enough in learning deformation-invariant global features which is essential in gastrointestinal endoscopic image classification. To solve this problem, in this paper we present a two-stage endoscopic image classification method which can effectively combine complementary advantages of midlevel CNN features and a capsule network. Specifically, the core of our method is a lesion-aware CNN feature extraction module which can encode sufficiently detailed information of lesions in midlevel CNN features and in turn enable the subsequent capsule classification network to effectively learn deformation-invariant relationships between image entities. Extensive experiments demonstrate the superiority of our method to the state-of-the-art methods with the classification accuracy of 94.83% on the Kvasir v2 data set and the classification accuracy of 85.99% on the HyperKvasir data set.

We propose iFlowGAN that learns an invertible flow (a sequence of invertible mappings) via adversarial learning and exploit it to transform a source distribution into a target distribution for unsupervised image-to-image translation. Existing GAN-based generative model such as CycleGAN [1], StarGAN [2], AGGAN [3] and CyCADA [4] needs to learn a highly under-constraint forward mapping F: X → Y from a source domain X to a target domain Y. Researchers do this by assuming there is a backward mapping B: Y → X such that x and y are fixed points of the composite functions B °F and F °B. Inspired by zero-order reverse filtering [5], we (1) understand F via contraction mappings on a metric space; (2) provide a simple yet effective algorithm to present B via the parameters of F in light of Banach fixed point theorem; (3) provide a Lipschitz-regularized network which indicates a general approach to compose the inverse for arbitrary Lipschitz-regularized networks via Banach fixed point theorem. This network is useful for image-to-image translation tasks because it could save the memory for the weights of B. Although memory can also be saved by directly coupling the weights of the forward and backward mappings, the performance of the image-to-image translation network degrades significantly. This explains why current GAN-based generative models including CycleGAN must take different parameters to compose the forward and backward mappings instead of employing the same weights to build both mappings. Taking advantage of the Lipschitz-regularized network, we not only build iFlowGAN to solve the redundancy shortcoming of CycleGAN but also assemble the corresponding iFlowGAN versions of StarGAN, AGGAN and CyCADA without breaking their network architectures. Extensive experiments show that the iFlowGAN version could produce comparable results of the original implementation while saving half parameters.

Joint filtering mainly uses an additional guidance image as a prior and transfers its structures to the target image in the filtering process. Different from existing approaches that rely on local linear models or hand-designed objective functions to extract the structural information from the guidance image, we propose a new joint filtering method based on a spatially variant linear representation model (SVLRM), where the target image is linearly represented by the guidance image. However, learning SVLRMs for vision tasks is a highly ill-posed problem. To estimate the spatially variant linear representation coefficients, we develop an effective approach based on a deep convolutional neural network (CNN). As such, the proposed deep CNN (constrained by the SVLRM) is able to model the structural information of both the guidance and input images. We show that the proposed approach can be effectively applied to a variety of applications, including depth/RGB image upsampling and restoration, flash deblurring, natural image denoising, and scale-aware filtering. In addition, we show that the linear representation model can be extended to high-order representation models (e.g., quadratic and cubic polynomial representations). Extensive experimental results demonstrate that the proposed method performs favorably against the state-of-the-art methods that have been specifically designed for each task.

Contextual information has been shown to be powerful for semantic segmentation. This work proposes a novel Context-based Tandem Network (CTNet) by interactively exploring the spatial contextual information and the channel contextual information, which can discover the semantic context for semantic segmentation. Specifically, the Spatial Contextual Module (SCM) is leveraged to uncover the spatial contextual dependency between pixels by exploring the correlation between pixels and categories. Meanwhile, the Channel Contextual Module (CCM) is introduced to learn the semantic features including the semantic feature maps and class-specific features by modeling the long-term semantic dependence between channels. The learned semantic features are utilized as the prior knowledge to guide the learning of SCM, which can make SCM obtain more accurate long-range spatial dependency. Finally, to further improve the performance of the learned representations for semantic segmentation, the results of the two context modules are adaptively integrated to achieve better results. Extensive experiments are conducted on four widely-used datasets, i.e., PASCAL-Context, Cityscapes, ADE20K and PASCAL VOC2012. The results demonstrate the superior performance of the proposed CTNet by comparison with several state-of-the-art methods. The source code and models are available at https://github.com/syp2ysy/CTNet.

Human motion prediction aims to generate future motions based on the observed human motions. Witnessing the success of Recurrent Neural Networks (RNN) in modeling sequential data, recent works utilize RNNs to model human-skeleton motions on the observed motion sequence and predict future human motions. However, these methods disregard the existence of the spatial coherence among joints and the temporal evolution among skeletons, which reflects the crucial characteristics of human motions in spatiotemporal space. To this end, we propose a novel Skeleton-Joint Co-Attention Recurrent Neural Networks (SC-RNN) to capture the spatial coherence among joints, and the temporal evolution among skeletons simultaneously on a skeleton-joint co-attention feature map in spatiotemporal space. First, a skeleton-joint feature map is constructed as the representation of the observed motion sequence. Second, we design a new Skeleton-Joint Co-Attention (SCA) mechanism to dynamically learn a skeleton-joint co-attention feature map of this skeleton-joint feature map, which can refine the useful observed motion information to predict one future motion. Third, a variant of GRU embedded with SCA collaboratively models the human-skeleton motion and human-joint motion in spatiotemporal space by regarding the skeleton-joint co-attention feature map as the motion context. Experimental results of human motion prediction demonstrate that the proposed method outperforms the competing methods.

This work aims to address the group activity recognition problem by exploring human motion characteristics. Traditional methods hold that the motions of all persons contribute equally to the group activity, which suppresses the contributions of some relevant motions to the whole activity while overstating some irrelevant motions. To address this problem, we present a Spatio-Temporal Context Coherence (STCC) constraint and a Global Context Coherence (GCC) constraint to capture the relevant motions and quantify their contributions to the group activity, respectively. Based on this, we propose a novel Coherence Constrained Graph LSTM (CCG-LSTM) with STCC and GCC to effectively recognize group activity, by modeling the relevant motions of individuals while suppressing the irrelevant motions. Specifically, to capture the relevant motions, we build the CCG-LSTM with a temporal confidence gate and a spatial confidence gate to control the memory state updating in terms of the temporally previous state and the spatially neighboring states, respectively. In addition, an attention mechanism is employed to quantify the contribution of a certain motion by measuring the consistency between itself and the whole activity at each time step. Finally, we conduct experiments on two widely-used datasets to illustrate the effectiveness of the proposed CCG-LSTM compared with the state-of-the-art methods.

Fine-grained image analysis (FGIA) is a longstanding and fundamental problem in computer vision and pattern recognition, and underpins a diverse set of real-world applications. The task of FGIA targets analyzing visual objects from subordinate categories, e.g., species of birds or models of cars. The small inter-class and large intra-class variation inherent to fine-grained image analysis makes it a challenging problem. Capitalizing on advances in deep learning, in recent years we have witnessed remarkable progress in deep learning powered FGIA. In this paper we present a systematic survey of these advances, where we attempt to re-define and broaden the field of FGIA by consolidating two fundamental fine-grained research areas - fine-grained image recognition and fine-grained image retrieval. In addition, we also review other key issues of FGIA, such as publicly available benchmark datasets and related domain-specific applications. We conclude by highlighting several research directions and open problems which need further exploration from the community.

Advances in facial manipulation technology have led to increasing indistinguishable and realistic face swap videos, which raises growing concerns about the security risk of deepfakes in the community. Although current deepfake detectors can gain promising performance when handling high-quality faces under within-database settings, most detectors suffer from performance degradation in cross-database evaluation. Moreover, when test faces’ quality is different from training faces, the performance degrades even under within-database settings. To this end, we propose a novel Localization invariance Siamese Network (LiSiam) to enforce localization invariance against different image degradation for deepfake detection. Specifically, our Siamese network-based feature extractor takes the original image and the corresponding quality-degraded image as pairwise inputs and outputs two segmentation maps. A localization invariance loss is further proposed to impose localization consistency between the two segmentation maps. In addition, we design a Mask-guided Transformer to capture the co-occurrence between the forgery region and its surroundings. Finally, a multi-task learning strategy is utilized to obtain a robust and discriminative feature representation and jointly optimize multiple objective functions (i.e., segmentation, classification, and localization invariance losses) in an end-to-end manner. Experimental results on two public datasets, i.e., FaceForensics++ and Celeb-DF, demonstrate the superior performance of our proposed method to state-of-the-art methods.

We propose an effective image dehazing algorithm which explores useful information from the input hazy image itself as the guidance for the haze removal. The proposed algorithm first uses a deep pre-dehazer to generate an intermediate result, and takes it as the reference image due to the clear structures it contains. To better explore the guidance information in the generated reference image, it then develops a progressive feature fusion module to fuse the features of the hazy image and the reference image. Finally, the image restoration module takes the fused features as input to use the guidance information for better clear image restoration. All the proposed modules are trained in an end-to-end fashion, and we show that the proposed deep pre-dehazer with progressive feature fusion module is able to help haze removal. Extensive experimental results show that the proposed algorithm performs favorably against state-of-the-art methods on the widely-used dehazing benchmark datasets as well as real-world hazy images.

Recent video frame interpolation methods have employed the curvilinear motion model to accommodate nonlinear motion among frames. The effectiveness of such model often hinges on motion estimation and occlusion detection, and therefore is greatly challenged when these methods are used to handle dynamic scenes that contain complex motions and occlusions. We address the challenges by proposing a bi-directional pseudo-three-dimensional network to exploit the correlation between motion estimation and depth-related occlusion estimation that considers the third dimension: depth. Specifically, the network exploits the correlation by learning shared multi-scale spatiotemporal representations, and by coupling the estimations, in both the past and future directions, to synthesize intermediate frames through a bi-directional pseudo-three-dimensional warping layer, where adaptive convolution kernels are estimated progressively from the coalescence of motion and depth-related occlusion estimations across multiple scales to acquire nonlocal and adaptive neighborhoods. The proposed network utilizes a novel multi-task collaborative learning strategy, which facilitates the supervised learning of video frame interpolation using complementary self-supervisory signals from motion and depth-related occlusion estimations. Across various benchmark datasets, the proposed method outperforms state-of-the-art methods in terms of accuracy, model size and runtime performance.

How to explore useful information from depth is the key success of the RGB-D saliency detection methods. While the RGB and depth images are from different domains, a modality gap will lead to unsatisfactory results for simple feature concatenation. Towards better performance, most methods focus on bridging this gap and designing different cross-modal fusion modules for features, while ignoring explicitly extracting some useful consistent information from them. To overcome this problem, we develop a simple yet effective RGB-D saliency detection method by learning discriminative cross-modality features based on the deep neural network. The proposed method first learns modality-specific features for RGB and depth inputs. And then we separately calculate the correlations of every pixel-pair in a cross-modality consistent way, i.e., the distribution ranges are consistent for the correlations calculated based on features extracted from RGB (RGB correlation) or depth inputs (depth correlation). From different perspectives, color or spatial, the RGB and depth correlations end up at the same point to depict how tightly each pixel-pair is related. Secondly, to complemently gather RGB and depth information, we propose a novel correlation-fusion to fuse RGB and depth correlations, resulting in a cross-modality correlation. Finally, the features are refined with both long-range cross-modality correlations and local depth correlations to predict salient maps. In which, the long-range cross-modality correlation provides context information for accurate localization, and the local depth correlation keeps good subtle structures for fine segmentation. In addition, a lightweight DepthNet is designed for efficient depth feature extraction. We solve the proposed network in an end-to-end manner. Both quantitative and qualitative experimental results demonstrate the proposed algorithm achieves favorable performance against state-of-the-art methods.

Fine-grained image hashing is challenging due to the difficulties of capturing discriminative local information to generate hash codes. On the one hand, existing methods usually extract local features with the dense attention mechanism by focusing on dense local regions, which cannot contain diverse local information for fine-grained hashing. On the other hand, hash codes of the same class suffer from large intra-class variation of fine-grained images. To address the above problems, this work proposes a novel sub-Region Localized Hashing (sRLH) to learn intra-class compact and inter-class separable hash codes that also contain diverse subtle local information for efficient fine-grained image retrieval. Specifically, to localize diverse local regions, a sub-region localization module is developed to learn discriminative local features by locating the peaks of non-overlap sub-regions in the feature map. Different from localizing dense local regions, these peaks can guide the sub-region localization module to capture multifarious local discriminative information by paying close attention to dispersive local regions. To mitigate intra-class variations, hash codes of the same class are enforced to approach one common binary center. Meanwhile, the gram-schmidt orthogonalization is performed on the binary centers to make the hash codes inter-class separable. Extensive experimental results on four widely used fine-grained image retrieval datasets demonstrate the superiority of sRLH to several state-of-the-art methods. The source code of sRLH will be released at https://github.com/ZhangYajie-NJUST/sRLH.git.

Visual odometry (VO) is a fundamental technique for many robotics and augmented reality (AR) applications. However, most existing RGB-D VO systems suffer from large performance degradation when large occlusions are present and/or a large portion of depth values are invalid due to the limited range of an RGB-D camera, prohibiting the usage of most systems in practical applications. To address above two problems, we present RGB-D DSO, an RGB-D direct sparse odometry with the core part being sliding-window optimization with occlusion removal and a depth refinement module. Occlusion removal excludes negative effects arising from occluded objects when minimizing the final energy function for camera pose tracking. Depth refinement ensures sufficient valid depth values uniformly distributed for the depth map of a keyframe. Experimental results on three public datasets demonstrate that our method achieves smaller tracking error than most existing state-of-the-art methods. Meanwhile, our system takes only 21.93 ms to track a frame, which is faster than most existing methods.

Group activity recognition (GAR) aiming at understanding the behavior of a group of people in a video clip has received increasing attention recently. Nevertheless, most of the existing solutions ignore that not all the persons contribute to the group activity of the scene equally. That is to say, the contribution from different individual behaviors to group activity is different; meanwhile, the contribution from people with different spatial positions is also different. To this end, we propose a novel Position-aware Participation-Contributed Temporal Dynamic Model (P2CTDM), in which two types of the key actor are constructed and learned. Specifically, we focus on the behaviors of key actors, who maintain steady motions (long moving time, called long motions) or display remarkable motions (but closely related to other people and the group activity, called flash motions) at a certain moment. For capturing long motions, we rank individual motions according to their intensity measured by stacking optical flows. For capturing flash motions that are closely related to other people, we design a position-aware interaction module (PIM) that simultaneously considers the feature similarity and position information. Beyond that, for capturing flash motions that are highly related to the group activity, we also present an aggregation long short-term memory (Agg-LSTM) to fuse the outputs from PIM by time-varying trainable attention factors. Four widely used benchmarks are adopted to evaluate the performance of the proposed P2CTDM compared to the state of the art.

Recently, many Visual Question Answering (VQA) models rely on the correlations between questions and answers yet neglect those between the visual information and the textual information. They would perform badly if the handled data distribute differently from the training data (i.e., out-of-distribution (OOD) data). Towards this end, we propose a two-stage unbiased VQA approach that addresses the unbiased issue from a causal perspective. In the causal inference stage, we mark the spurious correlation on the causal graph, explore the counterfactual causality, and devise a causal target based on the inherent correlations between the conventional and counterfactual VQA models. In the distillation stage, we introduce the causal target into the training process and leverages distilling as well as curriculum learning to capture the unbiased model. Since Causal Inference with Knowledge Distilling and Curriculum Learning (CKCL) reinforces the contribution of the visual information and eliminates the impact of the spurious correlation by distilling the knowledge in causal inference to the VQA model, it contributes to the good performance on both the standard data and out-of-distribution data. The extensive experimental results on VQA-CP v2 dataset demonstrate the superior performance of the proposed method compared to the state-of-the-art (SotA) methods.

Localization Quality Estimation (LQE) is crucial and popular in the recent advancement of dense object detectors since it can provide accurate ranking scores that benefit the Non-Maximum Suppression processing and improve detection performance. As a common practice, most existing methods predict LQE scores through vanilla convolutional features shared with object classification or bounding box regression. In this paper, we explore a completely novel and different perspective to perform LQE – based on the learned distributions of the four parameters of the bounding box. The bounding box distributions are inspired and introduced as "General Distribution" in GFLV1, which describes the uncertainty of the predicted bounding boxes well. Such a property makes the distribution statistics of a bounding box highly correlated to its real localization quality. Specifically, a bounding box distribution with a sharp peak usually corresponds to high localization quality, and vice versa. By leveraging the close correlation between distribution statistics and the real localization quality, we develop a considerably lightweight Distribution-Guided Quality Predictor (DGQP) for reliable LQE based on GFLV1, thus producing GFLV2. To our best knowledge, it is the first attempt in object detection to use a highly relevant, statistical representation to facilitate LQE. Extensive experiments demonstrate the effectiveness of our method. Notably, GFLV2 (ResNet101) achieves 46.2 AP at 14.6 FPS, surpassing the previous state-of-the-art ATSS baseline (43.6 AP at 14.6 FPS) by absolute 2.6 AP on COCO test-dev, without sacrificing the efficiency both in training and inference.

OCR-based image captioning aims to automatically describe images based on all the visual entities (both visual objects and scene text) in images. Compared with conventional image captioning, the reasoning of scene text is required for OCR-based image captioning since the generated descriptions often contain multiple OCR tokens. Existing methods attempt to achieve this goal via encoding the OCR tokens with rich visual and semantic representations. However, strong correlations between OCR tokens may not be established with such limited representations. In this paper, we propose to enhance the connections between OCR tokens from the viewpoint of exploiting the geometrical relationship. We comprehensively consider the height, width, distance, IoU and orientation relations between the OCR tokens for constructing the geometrical relationship. To integrate the learned relation as well as the visual and semantic representations into a unified framework, a Long Short-Term Memory plus Relation-aware pointer network (LSTM-R) architecture is presented in this paper. Under the guidance of the geometrical relationship between OCR tokens, our LSTM-R capitalizes on a newly-devised relation-aware pointer network to select OCR tokens from the scene text for OCR-based image captioning. Extensive experiments demonstrate the effectiveness of our LSTM-R. More remarkably, LSTM-R achieves state-of-the-art performance on TextCaps, with the CIDEr-D score being increased from 98.0% to 109.3%.

Existing video super-resolution (SR) algorithms usually assume that the blur kernels in the degradation process are known and do not model the blur kernels in the restoration. However, this assumption does not hold for blind video SR and usually leads to over-smoothed super-resolved frames. In this paper, we propose an effective blind video SR algorithm based on deep convolutional neural networks (CNNs). Our algorithm first estimates blur kernels from low-resolution (LR) input videos. Then, with the estimated blur kernels, we develop an effective image deconvolution method based on the image formation model of blind video SR to generate intermediate latent frames so that sharp image contents can be restored well. To effectively explore the information from adjacent frames, we estimate the motion fields from LR input videos, extract features from LR videos by a feature extraction network, and warp the extracted features from LR inputs based on the motion fields. Moreover, we develop an effective sharp feature exploration method which first extracts sharp features from restored intermediate latent frames and then uses a transformation operation based on the extracted sharp features and warped features from LR inputs to generate better features for HR video restoration. We formulate the proposed algorithm into an end-to-end trainable framework and show that it performs favorably against state-of-the-art methods.

In this paper, we seek reasons for the two major failure cases in Semantic Segmentation (SS): 1) missing small objects or minor object parts, and 2) mislabeling minor parts of large objects as wrong classes. We have an interesting finding that Failure-1 is due to the underuse of detailed features and Failure-2 is due to the underuse of visual contexts. To help the model learn a better trade-off, we introduce several Self-Regulation (SR) losses for training SS neural networks. By "self", we mean that the losses are from the model per se without using any additional data or supervision. By applying the SR losses, the deep layer features are regulated by the shallow ones to preserve more details; meanwhile, shallow layer classification logits are regulated by the deep ones to capture more semantics. We conduct extensive experiments on both weakly and fully supervised SS tasks, and the results show that our approach consistently surpasses the baselines. We also validate that SR losses are easy to implement in various state-of-the-art SS models, e.g., SPGNet [7] and OCRNet [62], incurring little computational overhead during training and none for testing1.

Significant advances have been made in image deraining due to the use of kinds of deep neural networks. However, existing deep neural network-based methods usually contain significant abundant network parameters and thus leads to expensive computation cost, which limits the application of deraining technology in high-level vision tasks. In this paper, we propose a compact and flexible Tree-structured Channel-wise Refinement Block (TCRB) for efficient image deraining, which contains augmentation, refinement, and aggregation modules to better explore features. Specifically, the refinement module can progressively extract groups of more discriminative features from the channel augmented inputs, and then the aggregation module adaptively fuses features from the refinement module to preserve image details by leveraging the Enhanced Channel Attention (ECA) method. Moreover, we present a Tree-structured Channel-wise Refinement Network (TCRN) by stacking multiple TCRBs, which could achieve competitive performance as the complicated networks. We embed the TCRB into a Multi-scale Tree-structured Channel-wise Refinement Network (MTCRN) based on an encoder and decoder network architecture and show that it performs favorably against state-of-the-art deraining algorithms on both synthetic datasets and real-world rainy images, while reaching a better trade-off in terms of model parameters and inference time.

In this companion paper, we provide the details of the reproducibility artifacts of the paper "Visual Relation of Interest Detection" presented at MM'20. Visual Relation of Interest Detection (VROID) aims to detect visual relations that are important for conveying the main content of an image. In this paper, we explain the file structure of the source code and publish the details of our ViROI dataset, which can be used to retrain the model with custom parameters. We also detail the scripts for component analysis and comparison with other methods and list the parameters that can be modified for custom training and inference.

This paper presents an anomaly detection method that is based on a sparse coding inspired Deep Neural Networks (DNN). Specifically, in light of the success of sparse coding based anomaly detection, we propose a Temporally-coherent Sparse Coding (TSC), where a temporally-coherent term is used to preserve the similarity between two similar frames. The optimization of sparse coefficients in TSC with the Sequential Iterative Soft-Thresholding Algorithm (SIATA) is equivalent to a special stacked Recurrent Neural Networks (sRNN) architecture. Further, to reduce the computational cost in alternatively updating the dictionary and sparse coefficients in TSC optimization and to alleviate hyperparameters selection in TSC, we stack one more layer on top of the TSC-inspired sRNN to reconstruct the inputs, and arrive at an sRNN-AE. We further improve sRNN-AE in the following aspects: i) rather than using a predefined similarity measurement between two frames, we propose to learn a data-dependent similarity measurement between neighboring frames in sRNN-AE to make it more suitable for anomaly detection; ii) to reduce computational costs in the inference stage, we reduce the depth of the sRNN in sRNN-AE and, consequently, our framework achieves real-time anomaly detection; iii) to improve computational efficiency, we conduct temporal pooling over the appearance features of several consecutive frames for summarizing information temporally, then we feed appearance features and temporally summarized features into a separate sRNN-AE for more robust anomaly detection. To facilitate anomaly detection evaluation, we also build a large-scale anomaly detection dataset which is even larger than the summation of all existing datasets for anomaly detection in terms of both the volume of data and the diversity of scenes. Extensive experiments on both a toy dataset under controlled settings and real datasets demonstrate that our method significantly outperforms existing methods, which validates the effectiveness of our sRNN-AE method for anomaly detection. Codes and data have been released at https://github.com/StevenLiuWen/sRNN_TSC_Anomaly_Detection.

We present an algorithm to directly solve numerous image restoration problems (e.g., image deblurring, image dehazing, and image deraining). These problems are ill-posed, and the common assumptions for existing methods are usually based on heuristic image priors. In this paper, we show that these problems can be solved by generative models with adversarial learning. However, a straightforward formulation based on a straightforward generative adversarial network (GAN) does not perform well in these tasks, and some structures of the estimated images are usually not preserved well. Motivated by an interesting observation that the estimated results should be consistent with the observed inputs under the physics models, we propose an algorithm that guides the estimation process of a specific task within the GAN framework. The proposed model is trained in an end-to-end fashion and can be applied to a variety of image restoration and low-level vision problems. Extensive experiments demonstrate that the proposed method performs favorably against state-of-the-art algorithms.

In this work, we aim to address the problem of human interaction recognition in videos by exploring the long-term inter-related dynamics among multiple persons. Recently, Long Short-Term Memory (LSTM) has become a popular choice to model individual dynamic for single-person action recognition due to its ability to capture the temporal motion information in a range. However, most existing LSTM-based methods focus only on capturing the dynamics of human interaction by simply combining all dynamics of individuals or modeling them as a whole. Such methods neglect the inter-related dynamics of how human interactions change over time. To this end, we propose a novel Hierarchical Long Short-Term Concurrent Memory (H-LSTCM) to model the long-term inter-related dynamics among a group of persons for recognizing human interactions. Specifically, we first feed each person's static features into a Single-Person LSTM to model the single-person dynamic. Subsequently, at one time step, the outputs of all Single-Person LSTM units are fed into a novel Concurrent LSTM (Co-LSTM) unit, which mainly consists of multiple sub-memory units, a new cell gate, and a new co-memory cell. In the Co-LSTM unit, each sub-memory unit stores individual motion information, while this Co-LSTM unit selectively integrates and stores inter-related motion information between multiple interacting persons from multiple sub-memory units via the cell gate and co-memory cell, respectively. Extensive experiments on several public datasets validate the effectiveness of the proposed H-LSTCM by comparing against baseline and state-of-the-art methods.

This article aims to tackle the problem of group activity recognition in the multiple-person scene. To model the group activity with multiple persons, most long short-term memory (LSTM)-based methods first learn the person-level action representations by several LSTMs and then integrate all the person-level action representations into the following LSTM to learn the group-level activity representation. This type of solution is a two-stage strategy, which neglects the "host-parasite" relationship between the group-level activity ("host") and person-level actions ("parasite") in spatiotemporal space. To this end, we propose a novel graph LSTM-in-LSTM (GLIL) for group activity recognition by modeling the person-level actions and the group-level activity simultaneously. GLIL is a "host-parasite" architecture, which can be seen as several person LSTMs (P-LSTMs) in the local view or a graph LSTM (G-LSTM) in the global view. Specifically, P-LSTMs model the person-level actions based on the interactions among persons. Meanwhile, G-LSTM models the group-level activity, where the person-level motion information in multiple P-LSTMs is selectively integrated and stored into G-LSTM based on their contributions to the inference of the group activity class. Furthermore, to use the person-level temporal features instead of the person-level static features as the input of GLIL, we introduce a residual LSTM with the residual connection to learn the person-level residual features, consisting of temporal features and static features. Experimental results on two public data sets illustrate the effectiveness of the proposed GLIL compared with state-of-the-art methods.

We present a simple and effective image super-resolution algorithm that imposes an image formation constraint on the deep neural networks via pixel substitution. The proposed algorithm first uses a deep neural network to estimate intermediate high-resolution images, blurs the intermediate images using known blur kernels, and then substitutes values of the pixels at the un-decimated positions with those of the corresponding pixels from the low-resolution images. The output of the pixel substitution process strictly satisfies the image formation model and is further refined by the same deep neural network in a cascaded manner. The proposed framework is trained in an end-to-end fashion and can work with existing feed-forward deep neural networks for super-resolution and converges fast in practice. Extensive experimental results show that the proposed algorithm performs favorably against state-of-the-art methods.

We present a simple and effective deep convolutional neural network (CNN) model for video deblurring. The proposed algorithm mainly consists of optical flow estimation from intermediate latent frames and latent frame restoration steps. It first develops a deep CNN model to estimate optical flow from intermediate latent frames and then restores the latent frames based on the estimated optical flow. To better explore the temporal information from videos, we develop a temporal sharpness prior to constrain the deep CNN model to help the latent frame restoration. We develop an effective cascaded training approach and jointly train the proposed CNN model in an end-to-end manner. We show that exploring the domain knowledge of video deblurring is able to make the deep CNN model more compact and efficient. Extensive experimental results show that the proposed algorithm performs favorably against state-of-the-art methods on the benchmark datasets as well as real-world videos.

The ability of recommending cold items (that have no behavior history) is a core strength of multimedia recommendation compared with behavior-only collaborative filtering. To learn effective item representation, a key challenge lies in the discrepancy between training and testing, since the cold items only exist in the testing data. This means that the signal used to represent an item varies during training and testing --- in the training stage, we can represent an item with both collaborative embedding and content embedding; whereas in the testing stage, we represent a cold item with content embedding only. Nevertheless, existing learning frameworks omit this critical discrepancy, resulting in suboptimal item representation for multimedia recommendation.

With the popularization of social websites, many methods have been proposed to explore the noisy tags for weakly-supervised image hashing.The main challenge lies in learning appropriate and sufficient information from those noisy tags. To address this issue, this work proposes a novel Masked visual-semantic Graph-based Reasoning Network, termed as MGRN, to learn joint visual-semantic representations for image hashing. Specifically, for each image, MGRN constructs a relation graph to capture the interactions among its associated tags and performs reasoning with Graph Attention Networks (GAT). MGRN randomly masks out one tag and then make GAT to predict this masked tag. This forces the GAT model to capture the dependence between the image and its associated tags, which can well address the problem of noisy tags. Thus it can capture key tags and visual structures from images to learn well-aligned visual-semantic representations. Finally, the auto-encoders is leveraged to learn hash codes that can preserve the local structure of the joint space. Meanwhile, the joint visual-semantic representations are reconstructed from those hash codes by using a decoder. Experimental results on two widely-used benchmark datasets demonstrate the superiority of the proposed method for image retrieval compared with several state-of-the-art methods.

Most metric-based meta-learning methods learn only the sophisticated similarity metric for few-shot classification, which may lead to the feature deterioration and unreliable prediction. Toward this end, we propose new mechanisms to learn generalized and discriminative feature embeddings as well as improve the robustness of classifiers against prediction corruptions for meta-learning. For this purpose, a new generation operator BlockMix is proposed by integrating interpolation on the images and labels within metric learning. Based on the above BlockMix, we propose a novel regularization method Meta Regularization as an auxiliary task branch with its own classifier to better constraint the feature embedding module and stabilize the meta-learning process. Furthermore, a novel inference scheme Self-Calibrated Inference is proposed to alleviate the unreliable prediction problem by calibrating the prototype of each category with the confidence-weighted average of the support and generated samples. The proposed mechanisms can be used as supplementary techniques alongside standard metric-based meta-learning algorithms without any pre-training. Experimental results demonstrate the insights and the efficiency of the proposed mechanisms respectively, compared with the state-of-the-art methods on the prevalent few-shot benchmarks.

OCR-based image captioning is the task of automatically describing images based on reading and understanding written text contained in images. Compared to conventional image captioning, this task is more challenging, especially when the image contains multiple text tokens and visual objects. The difficulties originate from how to make full use of the knowledge contained in the textual entities to facilitate sentence generation and how to predict a text token based on the limited information provided by the image. Such problems are not yet fully investigated in existing research. In this paper, we present a novel design - Multimodal Attention Captioner with OCR Spatial Relationship (dubbed as MMA-SR) architecture, which manages information from different modalities with a multimodal attention network and explores spatial relationships between text tokens for OCR-based image captioning. Specifically, the representations of text tokens and objects are fed into a three-layer LSTM captioner. Different attention scores for text tokens and objects are exploited through the multimodal attention network. Based on the attended features and the LSTM states, words are selected from the common vocabulary or from the image text by incorporating the learned spatial relationships between text tokens. Extensive experiments conducted on the TextCaps dataset verify the effectiveness of the proposed MMA-SR method. More remarkably, our MMA-SR increases CIDEr-D score from 93.7% to 98.0%.

In this paper, we propose a novel Visual Relation of Interest Detection (VROID) task, which aims to detect visual relations that are important for conveying the main content of an image, motivated from the intuition that not all correctly detected relations are really "interesting" in semantics and only a fraction of them really make sense for representing the image main content. Such relations are named Visual Relations of Interest (VROIs). VROID can be deemed as an evolution over the traditional Visual Relation Detection (VRD) task that tries to discover all visual relations in an image. We construct a new dataset to facilitate research on this new task, named ViROI, which contains 30,120 images each with VROIs annotated. Furthermore, we develop an Interest Propagation Network (IPNet) to solve VROID. IPNet contains a Panoptic Object Detection (POD) module, a Pair Interest Prediction (PaIP) module and a Predicate Interest Prediction (PrIP) module. The POD module extracts instances from the input image and also generates corresponding instance features and union features. The PaIP module then predicts the interest score of each instance pair while the PrIP module predicts that of each predicate for each instance pair. Then the interest scores of instance pairs are combined with those of the corresponding predicates as the final interest scores. All VROI candidates are sorted by final interest scores and the highest ones are taken as final results. We conduct extensive experiments to test effectiveness of our method, and the results show that IPNet achieves the best performance compared with the baselines on visual relation detection, scene graph generation and image captioning.

To support the replication of "Instance of Interest Detection", which was presented at MM'19, this companion paper provides the details of the artifacts. Instance of Interest Detection (IOID) aims to provide instance-level user interest modeling for image semantic description. In this paper, we explain the file structure of the source code and publish the details of our IOID dataset, which can be used to retrain the model with custom parameters. We also provide a program for component analysis to help other researchers to do experiments with alternative models that are not included in our experiments. Moreover, we provide a demo program for using our model easily.

Current Visual Question Answering (VQA) models mainly explore the statistical correlations between answers and questions, which fail to capture the relationship between the visual information and answers. The performance dramatically decreases when the distribution of handled data is different from the training data. Towards this end, this paper proposes a novel unbiased VQA model by exploring the Casual Inference with Knowledge Distillation (CIKD) to reduce the influence of bias. Specifically, the causal graph is first constructed to explore the counterfactual causality and infer the casual target based on the causal effect, which well reduces the bias from questions and obtain answers without training. Then knowledge distillation is leveraged to transfer the knowledge of the inferred casual target to the conventional VQA model. It makes the proposed method enable to handle both the biased data and standard data. To address the problem of the bad bias from the knowledge distillation, the ensemble learning is introduced based on the hypothetical bias reason. Experiments are conducted to show the performance of the proposed method. The significant improvements over the state-of-the-art methods on the VQA-CP v2 dataset well validate the contributions of this work.

One-stage detector basically formulates object detection as dense classification and localization (i.e., bounding box regression). The classification is usually optimized by Focal Loss and the box location is commonly learned under Dirac delta distribution. A recent trend for one-stage detectors is to introduce an \emph{individual} prediction branch to estimate the quality of localization, where the predicted quality facilitates the classification to improve detection performance. This paper delves into the \emph{representations} of the above three fundamental elements: quality estimation, classification and localization. Two problems are discovered in existing practices, including (1) the inconsistent usage of the quality estimation and classification between training and inference, and (2) the inflexible Dirac delta distribution for localization. To address the problems, we design new representations for these elements. Specifically, we merge the quality estimation into the class prediction vector to form a joint representation, and use a vector to represent arbitrary distribution of box locations. The improved representations eliminate the inconsistency risk and accurately depict the flexible distribution in real data, but contain \emph{continuous} labels, which is beyond the scope of Focal Loss. We then propose Generalized Focal Loss (GFL) that generalizes Focal Loss from its discrete form to the \emph{continuous} version for successful optimization. On COCO {\tt test-dev}, GFL achieves 45.0\% AP using ResNet-101 backbone, surpassing state-of-the-art SAPD (43.5\%) and ATSS (43.6\%) with higher or comparable inference speed.

We present a causal inference framework to improve Weakly-Supervised Semantic Segmentation (WSSS). Specifically, we aim to generate better pixel-level pseudo-masks by using only image-level labels -- the most crucial step in WSSS. We attribute the cause of the ambiguous boundaries of pseudo-masks to the confounding context, e.g., the correct image-level classification of "horse" and "person" may be not only due to the recognition of each instance, but also their co-occurrence context, making the model inspection (e.g., CAM) hard to distinguish between the boundaries. Inspired by this, we propose a structural causal model to analyze the causalities among images, contexts, and class labels. Based on it, we develop a new method: Context Adjustment (CONTA), to remove the confounding bias in image-level classification and thus provide better pseudo-masks as ground-truth for the subsequent segmentation model. On PASCAL VOC 2012 and MS-COCO, we show that CONTA boosts various popular WSSS methods to new state-of-the-arts.

The edge-preserving bilateral filter (BF) is a widely used smoothing tool in many applications. However, its brute-force implementation depends on the size of the box window. The shortcoming causes BF time-consuming for the image processing task with a large window. To make the computational complexity irrelevant to the window size, sparse approximations of the filtering kernels are calculated on a learned cosine dictionary by two steps. First, all possible frequencies are learned (estimated) from the filtering kernel to compose a cosine dictionary. Then, the sparse approximation is conducted on the learned dictionary to seek the optimal cosine approximation for both the range and spatial kernels. By making use of the one-dimensional cosine approximation for the range kernel, the BF is transformed into spatial convolutions. Subsequently, by employing the two-dimensional cosine approximation for the spatial kernel, spatial convolutions are decomposed into box filters of which the computational complexity is O(1). To the best of our knowledge, our approach is the first method that adaptively constructs a cosine dictionary according to the input kernel. This merit guarantees the best filtering accuracy and efficiency. These advantages are corroborated by several carefully designed experiments.

In feature learning tasks, one of the most enormous challenges is to generate an efficient discriminative subspace. In this paper, we propose a novel subspace learning method, named recursive discriminative subspace learning with an ℓ1-norm distance constraint (RDSL). RDSL can robustly extract features from the contaminated images and learn a discriminative subspace. With the use of an inequation-based ℓ1-norm distance metric constraint, the minimized ℓ1-norm distance metric objective function with slack variables induces samples in the same class to cluster as close as possible, meanwhile samples from different classes can be separated from each other as far as possible. By utilizing ℓ1-norm items in both the objective function and the constraint, RDSL can well handle the noisy data and outliers. In addition, the large margin formulation makes the proposed method insensitive to initializations. We describe two approaches to solve RDSL with a recursive strategy. Experimental results on six benchmark datasets, including the original data and the contaminated data, demonstrate that RDSL outperforms the state-of-the-art methods.

The existing research work on facial age synthesis has been mostly focused on long-term aging (e.g., over an age span of 10 years or more). In this paper, we employ generative adversarial networks (GANs) as a tool to investigate age synthesis over different age spans. Compared with long-term aging, short-term age synthesis suffers from the reduced amount of available training data, which can severely hinder the model training. We conduct a series of experiments to validate this. To facilitate short-term age synthesis, we further propose label distribution-guided generative adversarial network (ldGAN), where each sample is associated with an age label distribution (ALD) rather than a single age group. Accordingly, each sample can contribute not only to the learning of its own age group but also to neighbouring groups' learning. This is useful when addressing short-term aging to cope with the reduced amount of training data. In addition, unlike one-hot encoding which treats age groups as independent from one another, ldGAN can well capture the correlation among different age groups, so that smooth aging sequences can be achieved. The ALD model is integrated into GAN with a two-step process. Firstly, instead of the traditional one-hot encoding, ALD is applied as the condition of the generator. Secondly, we add a sequence of label distribution learners on top of several multi-scale discriminators, with the aim of minimizing the label distribution learning loss when optimizing both the generator and discriminators. Both qualitative and quantitative evaluations are conducted to assess ldGAN's ability in dealing with two core issues of face aging, i.e., aging effect generation and identity preservation. The obtained experimental results demonstrate the effectiveness of ldGAN in both learning short-term aging patterns and coping with the lack of training data.

Facial image synthesis has been extensively studied, for a long time, in both computer graphics and computer vision. Particularly, the synthesis of face images with varying ages, expressions and poses has received an increasing attention owing to several real-world applications. In this paper, facial age and expression synthesis are addressed. While previous and current research papers on facial age synthesis mostly adopt an age span of 10 years, this paper investigates face aging with a shorter time span. For expression synthesis, given a neutral face, we work on synthesizing faces with varying expression intensities (e.g., from zero to high). Note that both human ages and expression intensities are inherently ordinal. To fully exploit this ordinal nature, we devise ordinal ranking generative adversarial networks (ranking GAN). For each face, a one-hot label is assigned to define its age range/expression intensity. By exploiting the relative order information among age ranges/expression intensities, a binary ranking vector is further computed for each face. In ranking GAN, one-hot labels are used as the condition of the generator for synthesizing faces with target age groups/expression intensities. Moreover, we add a sequence of cost-sensitive ordinal rankers on top of several multi-scale discriminators, with the aim of minimizing age/intensity rank estimation loss when optimizing both the generator and discriminators. In order to evaluate the proposed ranking GAN, extensive experiments are carried out on several public face databases. As demonstrated by the experimental testing, this ranking scheme performs well even when the amount of available labeled training data is limited. The reported experimental results well demonstrate the effectiveness of ranking GAN on synthesizing face aging sequences and faces with varying expression intensities.

Haze interferes the transmission of scene radiation and significantly degrades color and details of outdoor images. Existing deep neural networks-based image dehazing algorithms usually use some common networks. The network design does not model the image formation of haze process well, which accordingly leads to dehazed images containing artifacts and haze residuals in some special scenes. In this paper, we propose a task-oriented network for image dehazing, where the network design is motivated by the image formation of haze process. The task-oriented network involves a hybrid network containing an encoder and decoder network and a spatially variant recurrent neural network which is derived from the hazy process. In addition, we develop a multi-stage dehazing algorithm to further improve the accuracy by filtering haze residuals in a step-bystep fashion. To constrain the proposed network, we develop a dual composition loss, content-based pixel-wise loss and total variation constraint. We train the proposed network in an end-to-end manner and analyze its effect on image dehazing. Experimental results demonstrate that the proposed algorithm achieves favorable performance against state-of-the-art dehazing methods.

With the prevalence of multimedia content on the Web, developing recommender solutions that can effectively leverage the rich signal in multimedia data is in urgent need. Owing to the success of deep neural networks in representation learning, recent advances on multimedia recommendation has largely focused on exploring deep learning methods to improve the recommendation accuracy. To date, however, there has been little effort to investigate the robustness of multimedia representation and its impact on the performance of multimedia recommendation. In this paper, we shed light on the robustness of multimedia recommender system. Using the state-of-the-art recommendation framework and deep image features, we demonstrate that the overall system is not robust, such that a small (but purposeful) perturbation on the input image will severely decrease the recommendation accuracy. This implies the possible weakness of multimedia recommender system in predicting user preference, and more importantly, the potential of improvement by enhancing its robustness. To this end, we propose a novel solution named Adversarial Multimedia Recommendation (AMR), which can lead to a more robust multimedia recommender model by using adversarial learning. The idea is to train the model to defend an adversary, which adds perturbations to the target image with the purpose of decreasing the model's accuracy. We conduct experiments on two representative multimedia recommendation tasks, namely, image recommendation and visually-aware product recommendation. Extensive results verify the positive effect of adversarial learning and demonstrate the effectiveness of our AMR method. Source codes are available in https://github.com/duxy-me/AMR.

Constrained Concept Factorization (CCF) yields the enhanced representation ability over CF by incorporating label information as additional constraints, but it cannot classify and group unlabeled data appropriately. Minimizing the difference between the original data and its reconstruction directly can enable CCF to model a small noisy perturbation, but is not robust to gross sparse errors. Besides, CCF cannot preserve the manifold structures in new representation space explicitly, especially in an adaptive manner. In this paper, we propose a joint label prediction based Robust Semi-Supervised Adaptive Concept Factorization (RS2ACF) framework. To obtain robust representation, RS2ACF relaxes the factorization to make it simultaneously stable to small entrywise noise and robust to sparse errors. To enrich prior knowledge to enhance the discrimination, RS2ACF clearly uses class information of labeled data and more importantly propagates it to unlabeled data by jointly learning an explicit label indicator for unlabeled data. By the label indicator, RS2ACF can ensure the unlabeled data of the same predicted label to be mapped into the same class in feature space. Besides, RS2ACF incorporates the joint neighborhood reconstruction error over the new representations and predicted labels of both labeled and unlabeled data, so the manifold structures can be preserved explicitly and adaptively in the representation space and label space at the same time. Owing to the adaptive manner, the tricky process of determining the neighborhood size or kernel width can be avoided. Extensive results on public databases verify that our RS2ACF can deliver state-of-the-art data representation, compared with other related methods.

Matrix factorization (MF) has been widely used to discover the low-rank structure and to predict the missing entries of data matrix. In many real-world learning systems, the data matrix can be very high dimensional but sparse. This poses an imbalanced learning problem since the scale of missing entries is usually much larger than that of the observed entries, but they cannot be ignored due to the valuable negative signal. For efficiency concern, existing work typically applies a uniform weight on missing entries to allow a fast learning algorithm. However, this simplification will decrease modeling fidelity, resulting in suboptimal performance for downstream applications. In this paper, we weight the missing data nonuniformly, and more generically, we allow any weighting strategy on the missing data. To address the efficiency challenge, we propose a fast learning method, for which the time complexity is determined by the number of observed entries in the data matrix rather than the matrix size. The key idea is twofold: 1) we apply truncated singular value decomposition on the weight matrix to get a more compact representation of the weights and 2) we learn MF parameters with elementwise alternating least squares (eALS) and memorize the key intermediate variables to avoid repeating computations that are unnecessary. We conduct extensive experiments on two recommendation benchmarks, demonstrating the correctness, efficiency, and effectiveness of our fast eALS method.

Joint filtering mainly uses an additional guidance image as a prior and transfers its structures to the target image in the filtering process. Different from existing algorithms that rely on locally linear models or hand-designed objective functions to extract the structural information from the guidance image, we propose a new joint filter based on a spatially variant linear representation model (SVLRM), where the target image is linearly represented by the guidance image. However, the SVLRM leads to a highly ill-posed problem. To estimate the linear representation coefficients, we develop an effective algorithm based on a deep convolutional neural network (CNN). The proposed deep CNN (constrained by the SVLRM) is able to estimate the spatially variant linear representation coefficients which are able to model the structural information of both the guidance and input images. We show that the proposed algorithm can be effectively applied to a variety of applications, including depth/RGB image upsampling and restoration, flash/no-flash image deblurring, natural image denoising, scale-aware filtering, etc. Extensive experimental results demonstrate that the proposed algorithm performs favorably against state-of-the-art methods that have been specially designed for each task.

Recent works attempt to improve scene parsing performance by exploring different levels of contexts, and typically train a well-designed convolutional network to exploit useful contexts across all pixels equally. However, in this paper, we find that the context demands are varying from different pixels or regions in each image. Based on this observation, we propose an Adaptive Context Network (ACNet) to capture the pixel-aware contexts by a competitive fusion of global context and local context according to different per-pixel demands. Specifically, when given a pixel, the global context demand is measured by the similarity between the global feature and its local feature, whose reverse value can also be used to measure the local context demand. We model the two demanding measurements by the proposed global context module and local context module, respectively, to generate their adaptive contextual features. Furthermore, we import multiple such modules to build several adaptive context blocks in different levels of network to obtain a coarse-to-fine result. Finally, comprehensive experimental evaluations demonstrate the effectiveness of the proposed ACNet, and new state-of-the-arts performances are achieved on all four public datasets, i.e. Cityscapes, ADE20K, PASCAL Context, and COCO Stuff.

Human can well recognize images of novel categories just after browsing few examples of these categories. One possible reason is that they have some external discriminative visual information about these categories from their prior knowledge. Inspired from this, we propose a novel Knowledge Transfer Network architecture (KTN) for few-shot image recognition. The proposed KTN model jointly incorporates visual feature learning, knowledge inferring and classifier learning into one unified framework for their optimal compatibility. First, the visual classifiers for novel categories are learned based on the convolutional neural network with the cosine similarity optimization. To fully explore the prior knowledge, a semantic-visual mapping network is then developed to conduct knowledge inference, which enables to infer the classifiers for novel categories from base categories. Finally, we design an adaptive fusion scheme to infer the desired classifiers by effectively integrating the above knowledge and visual information. Extensive experiments are conducted on two widely-used Mini-ImageNet and ImageNet Few-Shot benchmarks to evaluate the effectiveness of the proposed method. The results compared with the state-of-the-art approaches show the encouraging performance of the proposed method, especially on 1-shot and 2-shot tasks.

In this paper, we propose a road detection method with LiDAR-camera fusion in a novel conditional random field (CRF) framework to exploit both range and color information. In the LiDAR based part, a fast height-difference based scanning strategy is applied in the 2D LiDAR range-image domain and a dense road detection result in camera image domain can be obtained through geometric upsampling given the LiDAR-camera calibration parameters. In the camera based part, a fully convolutional network is applied in the camera image domain. Finally, we fuse the dense and binary road detection results from both LiDAR and camera in a single CRF framework. Experiments show that using a single thread of CPU, the proposed LiDAR based part can operate at a frequency of over 250Hz with sparse output in range image and 40Hz with dense result in camera image for the 64-beam Velodyne scanner. Our CRF fusion method achieves very promising road detection performance on the KITTI-Road dataset.

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%.

Most existing image enhancement methods function like a black box, which cannot clearly reveal the procedure behind each image enhancement operation. To overcome this limitation, in this paper, we design a progressive image enhancement framework, which generates an expected "good" retouched image with a group of self-interpretable image filters under the guidance of an aesthetic assessment model. The introduced aesthetic network effectively alleviates the shortage of paired training samples by providing extra supervision, and eliminate the bias caused by human subjective preferences. The self-interpretable image filters designed in our image enhancement framework, make the overall image enhancing procedure easy-to-understand. Extensive experiments demonstrate the effectiveness of our proposed framework.

Image captioning attempts to generate a sentence composed of several linguistic words, which are used to describe objects, attributes, and interactions in an image, denoted as visual semantic units in this paper. Based on this view, we propose to explicitly model the object interactions in semantics and geometry based on Graph Convolutional Networks (GCNs), and fully exploit the alignment between linguistic words and visual semantic units for image captioning. Particularly, we construct a semantic graph and a geometry graph, where each node corresponds to a visual semantic unit, i.e., an object, an attribute, or a semantic (geometrical) interaction between two objects. Accordingly, the semantic (geometrical) context-aware embeddings for each unit are obtained through the corresponding GCN learning processers. At each time step, a context gated attention module takes as inputs the embeddings of the visual semantic units and hierarchically align the current word with these units by first deciding which type of visual semantic unit (object, attribute, or interaction) the current word is about, and then finding the most correlated visual semantic units under this type. Extensive experiments are conducted on the challenging MS-COCO image captioning dataset, and superior results are reported when comparing to state-of-the-art approaches.

Acting as a bridge between vision and language, visual relationship detection (VRD) aims to represent objects and their interactions in an image with several relationship triplets. Nevertheless, the conventional VRD task shows little consideration for the penalization of incorrect relationship predictions, which in turn undermines its support for image understanding applications. In this paper, we propose a novel VRD task named hierarchical visual relationship detection (HVRD), which encourages predictions with abstract yet compatible relationship triplets when the confidence level of the specific image content is relatively low. Meanwhile, HVRD can handle the inevitable ambiguity of groundtruth annotation in VRD. Based on this, we propose a HVRD method, consisting of hierarchical object detection and hierarchical predicate detection. It can effectively detect the hierarchical visual relationships by exploiting both object concept hierarchy and predicate concept hierarchy with order embedding. We also propose the first datasets for HVRD evaluation, H-VRD and H-VG, by expanding the relationship category spaces of VRD and VG datasets to hierarchical ones respectively. The experimental results show that our method is superior to the state-of-the-art baselines.

Crowd counting aims to estimate the number of persons in a crowd image--a challenge until this day--as congestion degree varies, people's appearances may seem different. To address this problem, we propose a novel crowd counting method named Multi-layer Regression Network (MRNet), which consists of a multi-layer recognition branch and several density regressors. In practice, the recognition branch recognizes the congestion degree of the regions in a crowd image, then disintegrates the image into background and several crowd regions layer by layer, each regions are assigned different congestion degrees. In each layer, the recognized crowd regions with the specific congestion degree are delivered to a regressor with the corresponding density prior for crowd density estimation. The generated density maps at all layers are integrated to obtain the final density map for crowd density estimation. To date, MRNet is the first method to estimate crowd densities on crowd regions with different regressors. We conduct a comprehensive evaluation of MRNet on four typical datasets in comparison with nine state-of-the-art methods. By using multi-layer regression, MRNet achieves significant improvement in crowd counting accuracy, and outperforms the state-of-the-art methods.

In this paper, we propose a novel task named Instance of Interest Detection (IOID) to provide instance-level user interest modeling for image semantic description. IOID focuses on extracting the instances which are beneficial to represent image content, while other related tasks such as saliency analysis, attention model and instance segmentation extract the regions attracting visual attention or with a predefined category. To this end, we propose a Cross-influential Network for IOID, which integrates both visual saliency and semantic context. Moreover, we contribute the first dataset IOID evaluation, which consists of 45,000 images from MSCOCO with manually annotated instances of interest. Our method outperforms the state-of-the-art baselines on this dataset.

Mobile devices equipped with a monocular camera and an inertial measurement unit (IMU) are ideal platforms for augmented reality (AR) applications. However, nontrivial noises in low-cost IMUs, which are usually equipped in consumer-level mobile devices, could lead to large errors in pose estimation and in turn significantly degrade the user experience in mobile AR apps. In this study, we propose a novel monocular visual-inertial state estimation approach for robust and accurate pose estimation even for low-cost IMUs. The core of our method is an IMU pre-integration correction approach which effectively reduces the negative impact of IMU noises using the visual constraints in a sliding window and the kinematic constraint. We seamlessly integrate the IMU pre-integration correction module into a tightly-coupled,sliding-window based optimization framework for state estimation. Experimental results on public dataset EUROC demonstrate the superiority of our method to the state-of-the-art VINS-Mono in terms of smaller absolute trajectory errors (ATE) and relative pose errors (RPE). We further apply our method to real AR applications on two types of consumer-level mobile devices equipped with low-cost IMUs, i.e. an off-the-shelf smartphone and an AR glass. Experimental results demonstrate that our method can facilitate robust AR with little drifts on the two devices.

Image dehazing and deraining are import low-level compute vision tasks. In this paper, we propose a novel method named Selective Attention Network (SAN) to solve these two problems. Due to the density of haze and directions of rain streaks are complex and non-uniform, SAN adopts the channel-wise attention and spatial-channel attention to remove rain streaks and haze both in globally and locally. To better capture various of rain and hazy details, we propose a Selective Attention Module(SAM) to re-scale the channel-wise attention and spatial-channel attention instead of simple element-wise summation. In addition, we conduct ablation studies to validate the effectiveness of the each module of SAN. Extensive experimental results on synthetic and real-world datasets show that SAN performs favorably against state-of-the-art methods.

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.

Edge detection is a fundamental problem in computer vision. Recently, convolutional neural networks (CNNs) have pushed forward this field significantly. Existing methods which adopt specific layers of deep CNNs may fail to capture complex data structures caused by variations of scales and aspect ratios. In this paper, we propose an accurate edge detector using richer convolutional features (RCF). RCF encapsulates all convolutional features into more discriminative representation, which makes good usage of rich feature hierarchies, and is amenable to training via backpropagation. RCF fully exploits multiscale and multilevel information of objects to perform the image-to-image prediction holistically. Using VGG16 network, we achieve state-of-the-art performance on several available datasets. When evaluating on the well-known BSDS500 benchmark, we achieve ODS F-measure of 0.811 while retaining a fast speed (8 FPS). Besides, our fast version of RCF achieves ODS F-measure of 0.806 with 30 FPS. We also demonstrate the versatility of the proposed method by applying RCF edges for classical image segmentation.

Image retagging aims to improve the tag quality of social images by completing the missing tags, rectifying the noise-corrupted tags, and assigning new high-quality tags. Recent approaches simultaneously explore visual, user and tag information to improve the performance of image retagging by mining the tag-image-user associations. However, such methods will become computationally infeasible with the rapidly increasing number of images, tags and users. It has been proven that the anchor graph can significantly accelerate large-scale graph-based learning by exploring only a small number of anchor points. Inspired by this, we propose a novel Social anchor-Unit GrAph Regularized Tensor Completion (SUGAR-TC) method to efficiently refine the tags of social images, which is insensitive to the scale of data. First, we construct an anchor-unit graph across multiple domains (e.g., image and user domains) rather than traditional anchor graph in a single domain. Second, a tensor completion based on Social anchor-Unit GrAph Regularization (SUGAR) is implemented to refine the tags of the anchor images. Finally, we efficiently assign tags to non-anchor images by leveraging the relationship between the non-anchor units and the anchor units. Experimental results on a real-world social image database well demonstrate the effectiveness and efficiency of SUGAR-TC, outperforming the state-of-the-art methods.

The guided filter (GF) is a widely used smoothing tool in computer vision and image processing. However, to the best of our knowledge, few papers investigate the mathematical connection between this filter and the least-squares optimization. In this paper, we first interpret the guided filter as the cyclic coordinate descent (CCD) solver of a least-squares objective function. This discovery implies an extension approach to generalize the guided filter since we can change the least-squares objective function and define new filters as the first pass iteration of the CCD solver of modified objective functions. In addition, referring to the iterative minimizing procedure of the CCD, we can derive new rolling filtering schemes. So, we are reasonable to say that our discovery not only reveals an approach to design new GF-like filters adapting to specific requirements of applications but also offers thorough explanations for two rolling filtering schemes of the guided filter as well as the method to extend them. Experiments prove our new proposed filters and rolling filtering schemes could produce state-of-the-art results.

Hashing has attracted increasing research attention in recent years due to its high efficiency of computation and storage in image retrieval. Recent works have demonstrated the superiority of simultaneous feature representations and hash functions learning with deep neural networks. However, most existing deep hashing methods directly learn the hash functions by encoding the global semantic information, while ignoring the local spatial information of images. The loss of local spatial structure makes the performance bottleneck of hash functions, therefore limiting its application for accurate similarity retrieval. In this paper, we propose a novel deep ordinal hashing (DOH) method, which learns ordinal representations to generate ranking-based hash codes by leveraging the ranking structure of feature space from both local and global views. In particular, to effectively build the ranking structure, we propose to learn the rank correlation space by exploiting the local spatial information from fully convolutional network and the global semantic information from the convolutional neural network simultaneously. More specifically, an effective spatial attention model is designed to capture the local spatial information by selectively learning well-specified locations closely related to target objects. In such hashing framework, the local spatial and global semantic nature of images is captured in an end-to-end ranking-to-hashing manner. Experimental results conducted on three widely used datasets demonstrate that the proposed DOH method significantly outperforms the state-of-the-art hashing methods.

The use of multichannel data in line spectral estimation (or frequency estimation) is common for improving the estimation accuracy in array processing, structural health monitoring, wireless communications, and more. Recently proposed atomic norm methods have attracted considerable attention due to their provable superiority in accuracy, flexibility, and robustness compared with conventional approaches. In this paper, we analyze atomic norm minimization for multichannel frequency estimation from noiseless compressive data, showing that the sample size per channel that ensures exact estimation decreases with the increase of the number of channels under mild conditions. In particular, given L channels, order K (log K) (1 + L/1 log N) samples per channel, selected randomly from N equispaced samples, suffice to ensure with high probability exact estimation of K frequencies that are normalized and mutually separated by at least 4/N. Numerical results are provided corroborating our analysis.

Cross-modal hashing has attracted increasing research attention due to its efficiency for large-scale multimedia retrieval. With simultaneous feature representation and hash function learning, deep cross-modal hashing (DCMH) methods have shown superior performance. However, most existing methods on DCMH adopt binary quantization functions (e.g., [Formula: see text]) to generate hash codes, which limit the retrieval performance since binary quantization functions are sensitive to the variations of numeric values. Toward this end, we propose a novel end-to-end ranking-based hashing framework, in this paper, termed as deep semantic-preserving ordinal hashing (DSPOH), to learn hash functions with deep neural networks by exploring the ranking structure of feature dimensions. In DSPOH, the ordinal representation, which encodes the relative rank ordering of feature dimensions, is explored to generate hash codes. Such ordinal embedding benefits from the numeric stability of rank correlation measures. To make the hash codes discriminative, the ordinal representation is expected to well predict the class labels so that the ranking-based hash function learning is optimally compatible with the label predicting. Meanwhile, the intermodality similarity is preserved to guarantee that the hash codes of different modalities are consistent. Importantly, DSPOH can be effectively integrated with different types of network architectures, which demonstrates the flexibility and scalability of our proposed hashing framework. Extensive experiments on three widely used multimodal data sets show that DSPOH outperforms state of the art for cross-modal retrieval tasks.

Face identification (FI) via regression-based classification has been extensively studied during the recent years. Most vector-based methods achieve appealing performance in handing the noncontiguous pixelwise noises, while some matrix-based regression methods show great potential in dealing with contiguous imagewise noises. However, there is a lack of consideration of the mixture noises case, where both contiguous and noncontiguous noises are jointly contained. In this paper, we propose a weighted mixed-norm regression (WMNR) method to cope with the mixture image corruption. WMNR reveals certain essential characteristics of FI problems and bridges the vector- and matrix-based methods. Particularly, WMNR provides two advantages for both theoretical analysis and practical implementation. First, it generalizes possible distributions of the residuals into a unified feature weighted loss function. Second, it constrains the residual image as low-rank structure that can be quantified with general nonconvex functions and a weight factor. Moreover, a new reweighted alternating direction method of multipliers algorithm is derived for the proposed WMNR model. The algorithm exhibits great computational efficiency since it divides the original optimization problem into certain subproblems with analytical solution or can be implemented in a parallel manner. Extensive experiments on several public face databases demonstrate the advantages of WMNR over the state-of-the-art regression-based approaches. More specifically, the WMNR achieves an appealing tradeoff between identification accuracy and computational efficiency. Compared with the pure vector-based methods, our approach achieves more than 10% performance improvement and saves more than 70% of runtime, especially in severe corruption scenarios. Compared with the pure matrix-based methods, although it requires slightly more computation time, the performance benefits are even larger; up to 20% improvement can be obtained.

As the core of recommender systems, collaborative filtering (CF) models the affinity between a user and an item from historical user-item interactions, such as clicks, purchases, and so on. Benefiting from the strong representation power, neural networks have recently revolutionized the recommendation research, setting up a new standard for CF. However, existing neural recommender models do not explicitly consider the correlations among embedding dimensions, making them less effective in modeling the interaction function between users and items. In this work, we emphasize on modeling the correlations among embedding dimensions in neural networks to pursue higher effectiveness for CF. We propose a novel and general neural collaborative filtering framework—namely, ConvNCF, which is featured with two designs: (1) applying outer product on user embedding and item embedding to explicitly model the pairwise correlations between embedding dimensions, and (2) employing convolutional neural network above the outer product to learn the high-order correlations among embedding dimensions. To justify our proposal, we present three instantiations of ConvNCF by using different inputs to represent a user and conduct experiments on two real-world datasets. Extensive results verify the utility of modeling embedding dimension correlations with ConvNCF, which outperforms several competitive CF methods.

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.

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 present an algorithm to directly restore a clear image from a hazy image. This problem is highly ill-posed and most existing algorithms often use hand-crafted features, e.g., dark channel, color disparity, maximum contrast, to estimate transmission maps and then atmospheric lights. In contrast, we solve this problem based on a conditional generative adversarial network (cGAN), where the clear image is estimated by an end-to-end trainable neural network. Different from the generative network in basic cGAN, we propose an encoder and decoder architecture so that it can generate better results. To generate realistic clear images, we further modify the basic cGAN formulation by introducing the VGG features and an L1-regularized gradient prior. We also synthesize a hazy dataset including indoor and outdoor scenes to train and evaluate the proposed algorithm. Extensive experimental results demonstrate that the proposed method performs favorably against the state-of-the-art methods on both synthetic dataset and real world hazy images.

In this paper, we propose a general dual convolutional neural network (DualCNN) for low-level vision problems, e.g., super-resolution, edge-preserving filtering, deraining and dehazing. These problems usually involve the estimation of two components of the target signals: structures and details. Motivated by this, our proposed DualCNN consists of two parallel branches, which respectively recovers the structures and details in an end-to-end manner. The recovered structures and details can generate the target signals according to the formation model for each particular application. The DualCNN is a flexible framework for low-level vision tasks and can be easily incorporated into existing CNNs. Experimental results show that the DualCNN can be effectively applied to numerous low-level vision tasks with favorable performance against the state-of-the-art methods.

In this work, we contribute a new multi-layer neural network architecture named ONCF to perform collaborative filtering. The idea is to use an outer product to explicitly model the pairwise correlations between the dimensions of the embedding space. In contrast to existing neural recommender models that combine user embedding and item embedding via a simple concatenation or element-wise product, our proposal of using outer product above the embedding layer results in a two-dimensional interaction map that is more expressive and semantically plausible. Above the interaction map obtained by outer product, we propose to employ a convolutional neural network to learn high-order correlations among embedding dimensions. Extensive experiments on two public implicit feedback data demonstrate the effectiveness of our proposed ONCF framework, in particular, the positive effect of using outer product to model the correlations between embedding dimensions in the low level of multi-layer neural recommender model.

Recently, as an effective re-ranking technique, diffusion has attracted considerable attention in research on image retrieval. It inherits from random surfer model and is effective to deeply explore data manifold structure. However, as a common practice, diffusion is performed at query time which relies heavily on initial retrieval shortlists and suffers the bottleneck of online time-efficiency. To this end, in this paper, we present a more generalized method named CFA (cascaded feature augmentation) based on diffusion. First of all, we transfer diffusion process from online stage to offline stage and innovatively utilize output of diffusion to augment database features in a cascaded mode, which can eliminate iteration process at query time radically. Second, to scale the diffusion method to large image database, we propose a cascaded cluster diffusion technique for feature augmentation which largely reduces computational cost. Third, we extend our cascaded feature augmentation scheme to cases with multiple features without involving extra memory and time cost. Our CFA is compatible with other re-ranking methods. Extensive experiments on four public datasets demonstrate the effectiveness of our proposed algorithm.

With the development of deep neural networks, recent years have witnessed the increasing research interest on generative models. Specificly, Variational Auto-Encoders (VAE) and Generative Adversarial Networks (GAN) have achieved impressive results in various generative tasks. VAE is well established and theoretically elegant, but tends to generate blurry samples. In contrast, GAN has shown the advantage in visual quality of generated images, but suffers the difficulty in translating a random vector into a desired high-dimensional sample. As a result, the training dynamics in GAN are often unstable and the generated samples could collapse to limited modes. In this paper, we propose a new Auto-Encoder Generative Adversarial Networks (AEGAN), which takes advantages of both VAE and GAN. In our approach, instead of matching the encoded distribution of training samples to the prior Pz as in VAE, we map the random vector into the encoded latent space by adversarial training based on GAN. Besides, we also match the decoded distribution of training samples with that from random vectors. To evaluate our approach, we make comparison with other encoder-decoder based generative models on three public datasets. The experiments with both qualitative and quantitative results demonstrate the superiority of our algorithm over the comparison generative models.

Group activity recognition, a challenging task that a number of individuals occur in the scene of activity while only a small subset of them participate in, has received increasing attentions. However, most of the previous methods model all the individuals' actions equivalently while ignoring a fact that not all of them are contributed to the discrimination of group activity. That is to say, only a small number of key actors (participants) play important roles in the whole group activity. Inspired by this, we explore a new "One to Key" idea to progressively aggregate temporal dynamics of key actors with different participation degrees over time from each person. Here, we focus on two types of key actors in the whole activity, who steadily move in the whole process (long moving time) or intensely move (but closely related to the group activity) at a significant moment. Based on this, we propose a novel Participation-Contributed Temporal Dynamic Model (PC-TDM) to recognize group activity, which mainly consists of a "One" network and a "One to Key" network. Specifically, "One" network aims at modeling the individual dynamic of each person. "One to Key" network feeds the outputs from the "One" network into a Bidirectional LSTM (Bi-LSTM) according to the order of individual's moving time. Subsequently, each output state of Bi-LSTM weighted by a trainable time-varying attention factor is aggregated by going through LSTM one-by-one. Experimental results on two benchmarks demonstrate that the proposed method improves group activity recognition performance compared to the state-of-the-arts.

The high-dimensional convolution is widely used in various disciplines but has a serious performance problem due to its high computational complexity. Over the decades, people took a handmade approach to design fast algorithms for the Gaussian convolution. Recently, requirements for various non-Gaussian convolutions have emerged and are continuously getting higher. However, the handmade acceleration approach is no longer feasible for so many different convolutions since it is a time-consuming and painstaking job. Instead, we propose an Acceleration Network (AccNet) which turns the work of designing new fast algorithms to training the AccNet. This is done by: 1, interpreting splatting, blurring, slicing operations as convolutions; 2, turning these convolutions to $g$CP layers to build AccNet. After training, the activation function $g$ together with AccNet weights automatically define the new splatting, blurring and slicing operations. Experiments demonstrate AccNet is able to design acceleration algorithms for a ton of convolutions including Gaussian/non-Gaussian convolutions and produce state-of-the-art results.

Age progression is defined as aesthetically re-rendering the aging face at any future age for an individual face. In this work, we aim to automatically render aging faces in a personalized way. Basically, for each age group, we learn an aging dictionary to reveal its aging characteristics (e.g., wrinkles), where the dictionary bases corresponding to the same index yet from two neighboring aging dictionaries form a particular aging pattern cross these two age groups, and a linear combination of all these patterns expresses a particular personalized aging process. Moreover, two factors are taken into consideration in the dictionary learning process. First, beyond the aging dictionaries, each person may have extra personalized facial characteristics, e.g., mole, which are invariant in the aging process. Second, it is challenging or even impossible to collect faces of all age groups for a particular person, yet much easier and more practical to get face pairs from neighboring age groups. To this end, we propose a novel Bi-level Dictionary Learning based Personalized Age Progression (BDL-PAP) method. Here, bi-level dictionary learning is formulated to learn the aging dictionaries based on face pairs from neighboring age groups. Extensive experiments well demonstrate the advantages of the proposed BDL-PAP over other state-of-the-arts in term of personalized age progression, as well as the performance gain for cross-age face verification by synthesizing aging faces.

In this paper, we propose a novel fine-grained dictionary learning method for image classification. To learn a high-quality discriminative dictionary, three types of multispecific subdictionaries, i.e., class-specific dictionaries (CSDs), universal dictionary (UD), and family-specific dictionaries (FSDs), are simultaneously uncovered. Here, CSDs and UD, respectively, model the patterns for each class and the patterns irrespective of any class. FSDs can help reveal the shared patterns between multiple image classes, by filling the gap between the patterns in CSDs and UD. The dependence among image classes is revealed by the shared FSDs, and a common FSD can be assigned to several classes to represent their residual. Finally, the most discriminative FSD for each class is identified by minimizing the sparse reconstruction error. Extensive experiments are conducted on different widely used data sets for image classification. The results demonstrate the superior performance of the proposed method over some state-of-the-art methods.

Recent years have witnessed a dramatic increase in the number of community-contributed images. Hashing-based similarity searches for social images have been attracting considerable interest from computer vision and multimedia communities due to their computational and memory efficiency. In this paper, we propose a novel weakly supervised hashing method named weakly supervised multimodal hashing, for scalable social image retrieval. Semantic-aware hash functions are learned by jointly leveraging the weakly supervised tag information and visual information. Specifically, because user-provided tags associated with social images can describe the semantic information, the hash functions are learned by exploring the semantic structure. Unfortunately, the user-provided tags are imperfect. To avoid overfitting the weakly supervised tags, the local discriminative structure and the geometric structure in the visual space are explored. Besides, to learn compact and non-redundant hash codes, the hash functions are constrained to be orthogonal and an information theoretic regularization based on the maximum entropy principle is introduced to maximize the information provided by each hash code. The learned hash functions are orthogonal, which can avoid redundancy in the learned hash codes as much as possible. The proposed hashing learning problem is formulated as the eigenvalue problem, which can be solved efficiently. Extensive experiments are conducted on two widely used social image data sets and the encouraging performance compared with the state-of-the-art hashing techniques demonstrates the effectiveness of the proposed method.

Hashing methods have been widely used for approximate nearest neighbor search in recent years due to its computational and storage effectiveness. Most existing multimodal hashing methods try to preserve the similarity relationship based on either metric distances or semantic labels in a procrustean way, while ignoring the intra-class and inter-class variations inherent in the metric space. In this paper, we propose a novel multimodal hashing method, termed as semantic neighbor graph hashing (SNGH), which aims to preserve the fine-grained similarity metric based on the semantic graph that is constructed by jointly pursuing the semantic supervision and the local neighborhood structure. Specifically, the semantic graph is constructed to capture the local similarity structure for the image modality and the text modality, respectively. Furthermore, we define a function based on the local similarity in particular to adaptively calculate multi-level similarities by encoding the intra-class and inter-class variations. After obtaining the unified hash codes, the logistic regression with kernel trick is employed to learn view-specific hash functions independently for each modality. Extensive experiments are conducted on four widely used multimodal data sets. The experimental results demonstrate the superiority of the proposed SNGH method compared with the state-of-the-art multimodal hashing methods.

Videos are inherently multimodal. This paper studies the problem of exploiting the abundant multimodal clues for improved video classification performance. We introduce a novel hybrid deep learning framework that integrates useful clues from multiple modalities, including static spatial appearance information, motion patterns within a short time window, audio information, as well as long-range temporal dynamics. More specifically, we utilize three Convolutional Neural Networks (CNNs) operating on appearance, motion, and audio signals to extract their corresponding features. We then employ a feature fusion network to derive a unified representation with an aim to capture the relationships among features. Furthermore, to exploit the long-range temporal dynamics in videos, we apply two long short-term memory (LSTM) networks with extracted appearance and motion features as inputs. Finally, we also propose refining the prediction scores by leveraging contextual relationships among video semantics. The hybrid deep learning framework is able to exploit a comprehensive set of multimodal features for video classification. Through an extensive set of experiments, we demonstrate that: 1) LSTM networks that model sequences in an explicitly recurrent manner are highly complementary to the CNN models; 2) the feature fusion network that produces a fused representation through modeling feature relationships outperforms a large set of alternative fusion strategies; and 3) the semantic context of video classes can help further refine the predictions for improved performance. Experimental results on two challenging benchmarks-the UCF-101 and the Columbia Consumer Videos (CCV)-provide strong quantitative evidence that our framework can produce promising results: 93.1% on the UCF-101 and 84.5% on the CCV, outperforming several competing methods with clear margins.

Dimensionality reduction has attracted increasing attention, because high-dimensional data have arisen naturally in numerous domains in recent years. As one popular dimensionality reduction method, nonnegative matrix factorization (NMF), whose goal is to learn parts-based representations, has been widely studied and applied to various applications. In contrast to the previous approaches, this paper proposes a novel semisupervised NMF learning framework, called robust structured NMF, that learns a robust discriminative representation by leveraging the block-diagonal structure and the -norm (especially when ) loss function. Specifically, the problems of noise and outliers are well addressed by the -norm ( ) loss function, while the discriminative representations of both the labeled and unlabeled data are simultaneously learned by explicitly exploring the block-diagonal structure. The proposed problem is formulated as an optimization problem with a well-defined objective function solved by the proposed iterative algorithm. The convergence of the proposed optimization algorithm is analyzed both theoretically and empirically. In addition, we also discuss the relationships between the proposed method and some previous methods. Extensive experiments on both the synthetic and real-world data sets are conducted, and the experimental results demonstrate the effectiveness of the proposed method in comparison to the state-of-the-art methods.

This paper proposes a new discriminative deep quantization hashing (DDQH) approach for large-scale face image retrieval by learning discriminative and compact binary codes. It jointly explores the discrete code learning, batch normalization quantization (BNQ) module, and end-to-end learning in one unified framework, which can guarantee the optimal compatibility of hash coding and feature learning. To learn multiscale and robust facial features, a deep network properly stacking several convolution-pooling layers and pooling layers is designed, and the facial features are obtained by fusing the outputs of the last convolutional layer and the last pooling layer. Besides, the prediction errors of the learned binary codes are minimized to learn discriminative binary codes of images. To obtain higher retrieval accuracies, a BNQ module is utilized to control quantization at a moderate level. Experiments are conducted on two widely used data sets, and the proposed DDQH method achieves encouraging improvements over some state-of-the-art hashing approaches.

It has been well known that the user-provided tags of social images are imperfect, i.e., there exist noisy, irrelevant or incomplete tags. It heavily degrades the performance of many multimedia tasks. To alleviate this problem, we propose a Weakly-supervised Deep Nonnegative Low-rank model (WDNL) to improve the quality of tags by integrating the low-rank model with deep feature learning. A nonnegative low-rank model is introduced to uncover the intrinsic relationships between images and tags by simultaneously removing noisy or irrelevant tags and complementing missing tags. The deep architecture is leveraged to seamlessly connect the visual content and the semantic tag. That is, the proposed model can well handle the scalability by assigning tags to new images. Extensive experiments conducted on two real-world datasets demonstrate the effectiveness of the proposed method compared with some state-of-the-art methods.

The Guided Filter (GF) is well-known for its linear complexity. However, when filtering an image with an n-channel guidance, GF needs to invert an n × n matrix for each pixel. To the best of our knowledge existing matrix inverse algorithms are inefficient on current hardwares. This shortcoming limits applications of multichannel guidance in computation intensive system such as multi-label system. We need a new GF-like filter that can perform fast multichannel image guided filtering. Since the optimal linear complexity of GF cannot be minimized further, the only way thus is to bring all potentialities of current parallel computing hardwares into full play. In this paper we propose a hardware-efficient Guided Filter (HGF), which solves the efficiency problem of multichannel guided image filtering and yields competent results when applying it to multi-label problems with synthesized polynomial multichannel guidance. Specifically, in order to boost the filtering performance, HGF takes a new matrix inverse algorithm which only involves two hardware-efficient operations: element-wise arithmetic calculations and box filtering. In order to break the linear model restriction, HGF synthesizes a polynomial multichannel guidance to introduce nonlinearity. Benefiting from our polynomial guidance and hardware-efficient matrix inverse algorithm, HGF not only is more sensitive to the underlying structure of guidance but also achieves the fastest computing speed. Due to these merits, HGF obtains state-of-the-art results in terms of accuracy and efficiency in the computation intensive multi-label systems.

With the explosive growth of images containing faces, scalable face image retrieval has attracted increasing attention. Due to the amazing effectiveness, deep hashing has become a popular hashing method recently. In this work, we propose a new Discriminative Deep Hashing (DDH) network to learn discriminative and compact hash codes for large-scale face image retrieval. The proposed network incorporates the end-to-end learning, the divide-and-encode module and the desired discrete code learning into a unified framework. Specifically, a network with a stack of convolution-pooling layers is proposed to extract multi-scale and robust features by merging the outputs of the third max pooling layer and the fourth convolutional layer. To reduce the redundancy among hash codes and the network parameters simultaneously, a divide-and-encode module to generate compact hash codes. Moreover, a loss function is introduced to minimize the prediction errors of the learned hash codes, which can lead to discriminative hash codes. Extensive experiments on two datasets demonstrate that the proposed method achieves superior performance compared with some state-of-the-art hashing methods.

Convolutional Neural Networks (CNNs) have been widely used and achieve amazing performance, typically at the cost of very expensive computation. Some methods accelerate the CNN training by distributed GPUs those deploying GPUs on multiple servers. Unfortunately, they need to transmit a large amount of data among servers, which leads to long data transmitting time and long GPU idle time. Towards this end, we propose a novel hybrid parallelism architecture named "Wheel" to accelerate the CNN training by reducing the transmitted data and fully using GPUs simultaneously. Specifically, Wheel first partitions the layers of a CNN into two kinds of modules: convolutional module and fully-connected module, and deploys them following the proposed hybrid parallelism. In this way, Wheel transmits only a few parameters of CNNs among different servers, and transmits most of the parameters within the same server. The time to transmit data is significantly reduced. Second, to fully run each GPU and reduce the idle time, Wheel devises an alternate strategy deploying multiple workers on each GPU. Once one worker is suspended for receiving data, another one in the same GPU starts to execute the computing task. The workers in each GPU run concurrently and repeatedly like Wheels. Experiments are conducted to show the outperformance of the proposed scheme over the state-of-the-art parallel approaches.

Semantic attributes have been introduced as an effective representation for image classification especially in zero-shot learning. However, most of the existing semantic attributes are previously defined by people, thus the size of the attribute is restricted in practice and these attributes are not necessarily discriminative. Therefore, the classification accuracy is often relatively low using a fixed incomplete semantic attribute set for image representation. One intuitive solution is to expand the semantic attribute representation with some non-semantic features. However, how to make the supplementary features more effective and discriminative is still an open problem. In this paper, we propose a Discriminative Supplementary Feature Learning (DSFL) method to implement semantic attribute augmentation. In DSFL, the non-semantic supplementary features are learned simultaneously with the classifiers for the novel-categories. Extensive experiments are conducted on two public datasets and the results show that our approach achieves encouraging performance.

In this work, we address the human parsing task with a novel Contextualized Convolutional Neural Network (Co-CNN) architecture, which well integrates the cross-layer context, global image-level context, semantic edge context, within-super-pixel context and cross-super-pixel neighborhood context into a unified network. Given an input human image, Co-CNN produces the pixelwise categorization in an end-to-end way. First, the cross-layer context is captured by our basic local-to-global-to-local structure, which hierarchically combines the global semantic information and the local fine details across different convolutional layers. Second, the global image-level label prediction is used as an auxiliary objective in the intermediate layer of the Co-CNN, and its outputs are further used for guiding the feature learning in subsequent convolutional layers to leverage the global image-level context. Third, semantic edge context is further incorporated into Co-CNN, where the high-level semantic boundaries are leveraged to guide pixel-wise labeling. Finally, to further utilize the local super-pixel contexts, the within-super-pixel smoothing and cross-super-pixel neighbourhood voting are formulated as natural sub-components of the Co-CNN to achieve the local label consistency in both training and testing process. Comprehensive evaluations on two public datasets well demonstrate the significant superiority of our Co-CNN over other state-of-the-arts for human parsing. In particular, the F-1 score on the large dataset [1] reaches 81.72 percent by Co-CNN, significantly higher than 62.81 percent and 64.38 percent by the state-of-the-art algorithms, M-CNN [2] and ATR [1], respectively. By utilizing our newly collected large dataset for training, our Co-CNN can achieve 85.36 percent in F-1 score.

Social image tag refinement, which aims to improve tag quality by automatically completing the missing tags and rectifying the noise-corrupted ones, is an essential component for social image search. Conventional approaches mainly focus on exploring the visual and tag information, without considering the user information, which often reveals important hints on the (in)correct tags of social images. Towards this end, we propose a novel tri-clustered tensor completion framework to collaboratively explore these three kinds of information to improve the performance of social image tag refinement. Specifically, the inter-relations among users, images and tags are modeled by a tensor, and the intra-relations between users, images and tags are explored by three regularizations respectively. To address the challenges of the super-sparse and large-scale tensor factorization that demands expensive computing and memory cost, we propose a novel tri-clustering method to divide the tensor into a certain number of sub-tensors by simultaneously clustering users, images and tags into a bunch of tri-clusters. And then we investigate two strategies to complete these sub-tensors by considering (in)dependence between the sub-tensors. Experimental results on a real-world social image database demonstrate the superiority of the proposed method compared with the state-of-the-art methods.

Human action recognition has been well explored in applications of computer vision. Many successful action recognition methods have shown that action knowledge can be effectively learned from motion videos or still images. For the same action, the appropriate action knowledge learned from different types of media, e.g., videos or images, may be related. However, less effort has been made to improve the performance of action recognition in videos by adapting the action knowledge conveyed from images to videos. Most of the existing video action recognition methods suffer from the problem of lacking sufficient labeled training videos. In such cases, over-fitting would be a potential problem and the performance of action recognition is restrained. In this paper, we propose an adaptation method to enhance action recognition in videos by adapting knowledge from images. The adapted knowledge is utilized to learn the correlated action semantics by exploring the common components of both labeled videos and images. Meanwhile, we extend the adaptation method to a semi-supervised framework which can leverage both labeled and unlabeled videos. Thus, the over-fitting can be alleviated and the performance of action recognition is improved. Experiments on public benchmark datasets and real-world datasets show that our method outperforms several other state-of-the-art action recognition methods.

The number of images associated with weakly supervised user-provided tags has increased dramatically in recent years. User-provided tags are incomplete, subjective and noisy. In this paper, we focus on the problem of social image understanding, i.e., tag refinement, tag assignment, and image retrieval. Different from previous work, we propose a novel weakly supervised deep matrix factorization algorithm, which uncovers the latent image representations and tag representations embedded in the latent subspace by collaboratively exploring the weakly supervised tagging information, the visual structure, and the semantic structure. Due to the well-known semantic gap, the hidden representations of images are learned by a hierarchical model, which are progressively transformed from the visual feature space. It can naturally embed new images into the subspace using the learned deep architecture. The semantic and visual structures are jointly incorporated to learn a semantic subspace without overfitting the noisy, incomplete, or subjective tags. Besides, to remove the noisy or redundant visual features, a sparse model is imposed on the transformation matrix of the first layer in the deep architecture. Finally, a unified optimization problem with a well-defined objective function is developed to formulate the proposed problem and solved by a gradient descent procedure with curvilinear search. Extensive experiments on real-world social image databases are conducted on the tasks of image understanding: image tag refinement, assignment, and retrieval. Encouraging results are achieved with comparison with the state-of-the-art algorithms, which demonstrates the effectiveness of the proposed method.

Recent advances in multimedia ontology have resulted in a number of concept models, e.g., large-scale concept for multimedia and Mediamill 101, which are accessible and public to other researchers. However, most current research effort still focuses on building new concepts from scratch, very few work explores the appropriate method to construct new concepts upon the existing models already in the warehouse. To address this issue, we propose a new framework in this paper, termed LEarning Structured Model by Probabilistic loGic Ontology Tree for MultiM edia (LEGO1-MM), which can seamlessly integrate both the new target training examples and the existing primitive concept models to infer the more complex concept models. LEGO-MM treats the primitive concept models as the lego toy to potentially construct an unlimited vocabulary of new concepts. Specifically, we first formulate the logic operations to be the lego connectors to combine the existing concept models hierarchically in probabilistic logic ontology trees. Then, we incorporate new target training information simultaneously to efficiently disambiguate the underlying logic tree and correct the error propagation. Extensive experiments are conducted on a large vehicle domain data set from ImageNet. The results demonstrate that LEGO-MM has significantly superior performance over the existing state-of-the-art methods, which build new concept models from scratch.

This paper proposes a multi-grained random fields (MGRFs) model for mitosis identification. To deal with the difficulty in hidden state discovery and sequential structure modeling in mitosis sequences only containing gradual visual pattern changes, we design the graphical structure to transform individual sequence into a set of coarse-to-fine grained sequencesconveying diverse temporal dynamics. Furthermore, we propose the corresponding probabilistic model for joint temporal learning and feature learning. To deal with the non-convex formulation of MGRF, we decomposemodel training into two sub-tasks, layer-wise sequential learning of both temporal dynamics and visual feature and new layer generation by graph-based sequential grouping, and optimize the model by alternating between them iteratively. The proposed method is validated on very challenging mitosis data set of C3H10T1/2 and C2C12 stem cells. Extensive comparison experiments demonstrate its superiority to the state of the arts.

In this paper, we address the problem of searching for semantically similar images from a large database. We present a compact coding approach, supervised quantization. Our approach simultaneously learns feature selection that linearly transforms the database points into a low-dimensional discriminative subspace, and quantizes the data points in the transformed space. The optimization criterion is that the quantized points not only approximate the transformed points accurately, but also are semantically separable: the points belonging to a class lie in a cluster that is not overlapped with other clusters corresponding to other classes, which is formulated as a classification problem. The experiments on several standard datasets show the superiority of our approach over the state-of-the art supervised hashing and unsupervised quantization algorithms.

In this paper, we propose a robust probability based sparse method to solve single sample face recognition, which harvests the advantages of both local and global representation. Different from previous sparse representation methods that generate sparse coefficients by l1, we produce sparse class probability distribution by proposing a multi-phase sparse probability (MSP) framework. To create class probability distribution, we divide each face image into many local blocks and vote based on the classification results of all blocks. For classifying each block, we propose local similarity assumption that makes many conventional methods feasible to SSPP problem. Moreover, we also propose a heuristic multiphase class selection scheme to solve the entropy minimization problem, which finally provides a higher classification confidence from the global perspective. Experimental results on three popular databases show that our approach not only generalizes well to SSPP problem but also has strong robustness to expression, illumination, occlusion and time variation.

Recent years, with the development of e-commerce and population of mobile phones, image-based commodity retrieval has attracted much attention. This paper proposed a deep framework for commodity image retrieval(CMIR) from the view that they are same designed commodities. Our framework can catch as many design details as possible by exploring object detection and ranking sensitive feature learning, while the former is performed based on Faster R-CNN, and the later is learned with a multi-task Siamese Network. Besides, we refine the processing speed of the framework to make it a live system. Our framework is implemented on an android application based on Client/Server structure model whose server response time is about 150 ms per query.

The recently proposed low-rank representation (LRR) method has been empirically shown to be useful in various tasks such as motion segmentation, image segmentation, saliency detection and face recognition. While potentially powerful, LRR depends heavily on the configuration of its key parameter, λ. In realistic environments where the prior knowledge about data is lacking, however, it is still unknown how to choose λ in a suitable way. Even more, there is a lack of rigorous analysis about the success conditions of the method, and thus the significance of LRR is a little bit vague. In this paper we therefore establish a theoretical analysis for LRR, striving for figuring out under which conditions LRR can be successful, and deriving a moderately good estimate to the key parameter λ as well. Simulations on synthetic data points and experiments on real motion sequences verify our claims.

This paper presents a patch-set-based sparse representation for image set classification. Compared with image-based image set representation, our patch-set-based representation is alignment free and thus has an advantage for tasks like video-based face recognition, image-set-based object recognition, and video-based hand gesture recognition, where precious alignment is usually difficult or even impossible due to large variance in view angle or pose. Specifically, to bypass the alignment issue, we propose to adopt the patch-based image set representation by dividing each image within each set into patches, then we cluster all the training patches into multiple clusters and classify the test patches based on the cluster centers of training patches. The labels of test patches within each cluster are inferred from a patch-set-based sparse representation for classification, and the labels of all test patches from all the clusters are then aggregated to predict a single label for the test set. Experimental results on video-based face recognition data sets (CMU-MoBo and YouTube Celebrities), image-set-based object recognition data set (ETH-80), and video-based hand gesture recognition data set (Kinect Hand Gestures) demonstrate that our proposed method consistently outperforms all existing ones, and the improvement is very significant on the YouTube Celebrities and Kinect Hand Gesture data sets. Moreover, we also quantitatively show the robustness of our method to misalignment on the Mutli-PIE data set.

Capabilities of inference and prediction are the significant components of visual systems. Visual path prediction is an important and challenging task among them, with the goal to infer the future path of a visual object in a static scene. This task is complicated as it needs high-level semantic understandings of both the scenes and underlying motion patterns in video sequences. In practice, cluttered situations have also raised higher demands on the effectiveness and robustness of models. Motivated by these observations, we propose a deep learning framework, which simultaneously performs deep feature learning for visual representation in conjunction with spatiotemporal context modeling. After that, a unified path-planning scheme is proposed to make accurate path prediction based on the analytic results returned by the deep context models. The highly effective visual representation and deep context models ensure that our framework makes a deep semantic understanding of the scenes and motion patterns, consequently improving the performance on visual path prediction task. In experiments, we extensively evaluate the model's performance by constructing two large benchmark datasets from the adaptation of video tracking datasets. The qualitative and quantitative experimental results show that our approach outperforms the state-of-the-art approaches and owns a better generalization capability.

The nuclear norm is widely used as a convex surrogate of the rank function in compressive sensing for low rank matrix recovery with its applications in image recovery and signal processing. However, solving the nuclear norm-based relaxed convex problem usually leads to a suboptimal solution of the original rank minimization problem. In this paper, we propose to use a family of nonconvex surrogates of L0-norm on the singular values of a matrix to approximate the rank function. This leads to a nonconvex nonsmooth minimization problem. Then, we propose to solve the problem by an iteratively re-weighted nuclear norm (IRNN) algorithm. IRNN iteratively solves a weighted singular value thresholding problem, which has a closed form solution due to the special properties of the nonconvex surrogate functions. We also extend IRNN to solve the nonconvex problem with two or more blocks of variables. In theory, we prove that the IRNN decreases the objective function value monotonically, and any limit point is a stationary point. Extensive experiments on both synthesized data and real images demonstrate that IRNN enhances the low rank matrix recovery compared with the state-of-the-art convex algorithms.

Learning high-level image representations using object proposals has achieved remarkable success in multi-label image recognition. However, most object proposals provide merely coarse information about the objects, and only carefully selected proposals can be helpful for boosting the performance of multi-label image recognition. In this paper, we propose an object-proposal-free framework for multi-label image recognition: random crop pooling (RCP). Basically, RCP performs stochastic scaling and cropping over images before feeding them to a standard convolutional neural network, which works quite well with a max-pooling operation for recognizing the complex contents of multi-label images. To better fit the multi-label image recognition task, we further develop a new loss function-the dynamic weighted Euclidean loss-for the training of the deep network. Our RCP approach is amazingly simple yet effective. It can achieve significantly better image recognition performance than the approaches using object proposals. Moreover, our adapted network can be easily trained in an end-to-end manner. Extensive experiments are conducted on two representative multi-label image recognition data sets (i.e., PASCAL VOC 2007 and PASCAL VOC 2012), and the results clearly demonstrate the superiority of our approach.

It is a necessary but challenging task to relieve users from the proliferative news information and allow them to quickly and comprehensively master the information of the whats and hows that are happening in the world every day. In this article, we develop a novel approach of multimedia news summarization for searching results on the Internet, which uncovers the underlying topics among query-related news information and threads the news events within each topic to generate a query-related brief overview. First, the hierarchical latent Dirichlet allocation (hLDA) model is introduced to discover the hierarchical topic structure from query-related news documents, and a new approach based on the weighted aggregation and max pooling is proposed to identify one representative news article for each topic. One representative image is also selected to visualize each topic as a complement to the text information. Given the representative documents selected for each topic, a time-bias maximum spanning tree (MST) algorithm is proposed to thread them into a coherent and compact summary of their parent topic. Finally, we design a friendly interface to present users with the hierarchical summarization of their required news information. Extensive experiments conducted on a large-scale news dataset collected from multiple news Web sites demonstrate the encouraging performance of the proposed solution for news summarization in news retrieval.

In recent years, deep neural networks have been successfully applied to model visual concepts and have achieved competitive performance on many tasks. Despite their impressive performance, traditional deep networks are subjected to the decayed performance under the condition of lacking sufficient training data. This problem becomes extremely severe for deep networks trained on a very small dataset, making them overfitting by capturing nonessential or noisy information in the training set. Toward this end, we propose a novel generalized deep transfer networks (DTNs), capable of transferring label information across heterogeneous domains, textual domain to visual domain. The proposed framework has the ability to adequately mitigate the problem of insufficient training images by bringing in rich labels from the textual domain. Specifically, to share the labels between two domains, we build parameter- and representation-shared layers. They are able to generate domain-specific and shared interdomain features, making this architecture flexible and powerful in capturing complex information from different domains jointly. To evaluate the proposed method, we release a new dataset extended from NUS-WIDE at http://imag.njust.edu.cn/NUS-WIDE-128.html. Experimental results on this dataset show the superior performance of the proposed DTNs compared to existing state-of-the-art methods.

The quantization techniques have shown competitive performance in approximate nearest neighbor search. The state-of-the-art algorithm, composite quantization, takes advantage of the compositionabity, i.e., the vector approximation accuracy, as opposed to product quantization and Cartesian k-means. However, we have observed that the runtime cost of computing the distance table in composite quantization, which is used as a lookup table for fast distance computation, becomes nonnegligible in real applications, e.g., reordering the candidates retrieved from the inverted index when handling very large scale databases. To address this problem, we develop a novel approach, called sparse composite quantization, which constructs sparse dictionaries. The benefit is that the distance evaluation between the query and the dictionary element (a sparse vector) is accelerated using the efficient sparse vector operation, and thus the cost of distance table computation is reduced a lot. Experiment results on large scale ANN retrieval tasks (1M SIFTs and 1B SIFTs) and applications to object retrieval show that the proposed approach yields competitive performance: superior search accuracy to product quantization and Cartesian k-means with almost the same computing cost, and much faster ANN search than composite quantization with the same level of accuracy.

In this work, we address the human parsing task with a novel Contextualized Convolutional Neural Network (Co-CNN) architecture, which well integrates the cross-layer context, global image-level context, within-super-pixel context and cross-super-pixel neighborhood context into a unified network. Given an input human image, Co-CNN produces the pixel-wise categorization in an end-to-end way. First, the cross-layer context is captured by our basic local-to-global-to-local structure, which hierarchically combines the global semantic structure and the local fine details within the cross-layers. Second, the global image-level label prediction is used as an auxiliary objective in the intermediate layer of the Co-CNN, and its outputs are further used for guiding the feature learning in subsequent convolutional layers to leverage the global image-level context. Finally, to further utilize the local super-pixel contexts, the within-super-pixel smoothing and cross-super-pixel neighbourhood voting are formulated as natural sub-components of the Co-CNN to achieve the local label consistency in both training and testing process. Comprehensive evaluations on two public datasets well demonstrate the significant superiority of our Co-CNN architecture over other state-of-the-arts for human parsing. In particular, the F-1 score on the large dataset [15] reaches 76.95% by Co-CNN, significantly higher than 62.81% and 64.38% by the state-of-the-art algorithms, M-CNN [21] and ATR [15], respectively.

In this paper, we aim to automatically render aging faces in a personalized way. Basically, a set of age-group specific dictionaries are learned, where the dictionary bases corresponding to the same index yet from different dictionaries form a particular aging process pattern cross different age groups, and a linear combination of these patterns expresses a particular personalized aging process. Moreover, two factors are taken into consideration in the dictionary learning process. First, beyond the aging dictionaries, each subject may have extra personalized facial characteristics, e.g. mole, which are invariant in the aging process. Second, it is challenging or even impossible to collect faces of all age groups for a particular subject, yet much easier and more practical to get face pairs from neighboring age groups. Thus a personality-aware coupled reconstruction loss is utilized to learn the dictionaries based on face pairs from neighboring age groups. Extensive experiments well demonstrate the advantages of our proposed solution over other state-of-the-arts in term of personalized aging progression, as well as the performance gain for cross-age face verification by synthesizing aging faces.

With the proliferation of large-scale social images, recent years have witnessed the increasing amount of images with user-provided tags, which leads to considerable effort made on hashing based approximate nearest neighbor (ANN) search in huge databases. In this work, we propose a novel Semantic-aware Hashing method (SaH) by discovering knowledge from these social media resources to implement approximate similarity search. Different from the previous work, the proposed method learns semantic hashing codes by exploiting heterogeneous information from the textual and visual domains. The semantic structure in the textual domain is well preserved to learn the binary codes. To handle the noisy, incomplete, or subjective user-provided tags, the visual structure is also leveraged. On the other hand, an information theoretic regularization is exploited by using maximum entropy principle and a row-wise sparse model with l2,p (0 < p ≤ 1) mixed norm is introduced to filter certain noisy or redundant visual features. Experiments are conducted on a widely-used social image dataset and the comparison results demonstrate the outperforming performance of the proposed SaH method over state-of-the-art hashing techniques.

For the RGB-D object recognition task, the robust and rich representations can boost the performance. Most works employ feature learning approaches to learn specific representation for the RGB and depth modalities independently, while some directly learn common property. Different from them, this paper proposes a novel supervised feature learning method for RGB-D object recognition, named Partially Common-Semantic Learning (PCSL), which jointly captures the complementary and consistency semantic information from RGB and depth modalities. The complementary information is revealed by the individual modality, while the consistency is exploited by both modalities simultaneously. In PCSL, Reconstruction Independent Component Analysis (RICA) is extended to integrate the supervised information and learn both of the complementary and partially shared common semantic information. The proposed approach is evaluated on two public RGB-D datasets and achieves better performance than several state-of-the-art methods.

In recent years, deep networks have been successfully applied to model image concepts and achieved competitive performance on many data sets. In spite of impressive performance, the conventional deep networks can be subjected to the decayed performance if we have insufficient training examples. This problem becomes extremely severe for deep networks with powerful representation structure, making them prone to over fitting by capturing nonessential or noisy information in a small data set. In this paper, to address this challenge, we will develop a novel deep network structure, capable of transferring labeling information across heterogeneous domains, especially from text domain to image domain. This weakly-shared Deep Transfer Networks (DTNs) can adequately mitigate the problem of insufficient image training data by bringing in rich labels from the text domain.

"What shall I look like after N years?" In this paper, we present an Auto Age Progression system, which automatically renders a series of aging faces in the future age ranges and generates an aging sequence (aging video) covering the entire life for an individual input. In the offline stage, a set of age-range specific dictionaries are learned from the constructed database, where the dictionary bases corresponding to the same index yet from different dictionaries form a particular aging process pattern across different age groups, and a linear combination of these patterns expresses a particular personalized aging process. In the online stage, for an input face of an individual, our system renders the aging faces corresponding to different age ranges through the aging dictionaries, and then generates an age progression by the presented face morphing technology.

The number of images associated with user-provided tags has increased dramatically in recent years. User-provided tags are incomplete, subjective and noisy. In this work, we focus on the problem of image tag refinement and assignment. Different from previous work, we propose a novel Deep Matrix Factorization (DMF) algorithm, which uncovers the latent image representations and tag representations embedded in the latent subspace by exploiting the weakly-supervised tagging information and visual information. Due to the well-known semantic gap, the hidden representations of images are learned by a hierarchical model, which are progressively transformed from the visual feature space. It can naturally embed new images into the subspace using the learned deep architecture. Besides, to remove the noisy or redundant visual features, a sparse model is imposed on the transformation matrix of the first layer in the deep architecture. Finally, a unified optimization problem with a well-defined objective function is developed to formulate the proposed problem. Extensive experiments on real-world social image databases are conducted on the tasks of image tag refinement and assignment. Encouraging results are achieved with comparison to the state-of-the-art algorithms, which demonstrates the effectiveness of the proposed method.

To improve the performance of kinship verification, we propose a novel deep kinship verification (DKV) model by integrating excellent deep learning architecture into metric learning. Unlike most existing shallow models based on metric learning for kinship verification, we employ a deep learning model followed by a metric learning formulation to select nonlinear features, which can find the appropriate project space to ensure the margin of negative sample pairs (i.e. parent and child without kinship relation) as large as possible and the margin of positive sample pairs (i.e. parent and child with kinship relation) as small as possible. Experimental results show that our method achieves satisfactory performance on two widely-used benchmarks, i.e. KFW-I and KFW-II.

Satisfaction prediction is one of the prime concerns in search performance evaluation. It is a non-trivial task for two major reasons: (1) The definition of satisfaction is rather subjective and different users may have different opinions in satisfaction judgement. (2) Most existing studies on satisfaction prediction mainly rely on users' click-through or query reformulation behaviors but there are many sessions without such kind of interactions. To shed light on these research questions, we construct an experimental search engine that could collect users' satisfaction feedback as well as mouse click-through/movement data. Different from existing studies, we compare for the first time search users' and external assessors' opinions on satisfaction. We find that search users pay more attention to the utility of results while external assessors emphasize on the efforts spent in search sessions. Inspired by recent studies in predicting result relevance based on mouse movement patterns (namely motifs), we propose to estimate the utilities of search results and the efforts in search sessions with motifs extracted from mouse movement data on search result pages (SERPs). Besides the existing frequency-based motif selection method, two novel selection strategies (distance-based and distribution-based) are also adopted to extract high quality motifs for satisfaction prediction. Experimental results on over 1,000 user sessions show that the proposed strategies outperform existing methods and also have promising generalization capability for different users and queries.

To uncover an appropriate latent subspace for data representation, in this paper we propose a novel Robust Structured Subspace Learning (RSSL) algorithm by integrating image understanding and feature learning into a joint learning framework. The learned subspace is adopted as an intermediate space to reduce the semantic gap between the low-level visual features and the high-level semantics. To guarantee the subspace to be compact and discriminative, the intrinsic geometric structure of data, and the local and global structural consistencies over labels are exploited simultaneously in the proposed algorithm. Besides, we adopt the l2,1-norm for the formulations of loss function and regularization respectively to make our algorithm robust to the outliers and noise. An efficient algorithm is designed to solve the proposed optimization problem. It is noted that the proposed framework is a general one which can leverage several well-known algorithms as special cases and elucidate their intrinsic relationships. To validate the effectiveness of the proposed method, extensive experiments are conducted on diversity datasets for different image understanding tasks, i.e., image tagging, clustering, and classification, and the more encouraging results are achieved compared with some state-of-the-art approaches.

A robust and effective specular highlight removal method is proposed in this paper. It is based on a key observation--the maximum fraction of the diffuse colour component in diffuse local patches in colour images changes smoothly. The specular pixels can thus be treated as noise in this case. This property allows the specular highlights to be removed in an image denoising fashion: an edge-preserving low-pass filter (e.g., the bilateral filter) can be used to smooth the maximum fraction of the colour components of the original image to remove the noise contributed by the specular pixels. Recent developments in fast bilateral filtering techniques enable the proposed method to run over 200× faster than state-of-the-art techniques on a standard CPU and differentiates it from previous work.

Influence maximization, whose objective is to select k users (called seeds) from a social network such that the number of users influenced by the seeds (called influence spread) is maximized, has attracted significant attention due to its widespread applications, such as viral marketing and rumor control. However, in real-world social networks, users have their own interests (which can be represented as topics) and are more likely to be influenced by their friends (or friends' friends) with similar topics. We can increase the influence spread by taking into consideration topics. To address this problem, we study topic-aware influence maximization, which, given a topic-aware influence maximization (TIM) query, finds k seeds from a social network such that the topic-aware influence spread of the k seeds is maximized. Our goal is to enable online TIM queries. Since the topic-aware influence maximization problem is NP-hard, we focus on devising efficient algorithms to achieve instant performance while keeping a high influence spread. We utilize a maximum influence arborescence (MIA) model to approximate the computation of influence spread. To efficiently find k seeds under the MIA model, we first propose a best-effort algorithm with 1 − 1/ e approximation ratio, which estimates an upper bound of the topic-aware influence of each user and utilizes the bound to prune large numbers of users with small influence. We devise effective techniques to estimate tighter upper bounds. We then propose a faster topic-sample-based algorithm with ε · (1 − 1/ e ) approximation ratio for any ε ∈ (0, 1], which materializes the influence spread of some topic-distribution samples and utilizes the materialized information to avoid computing the actual influence of users with small influences. Experimental results show that our methods significantly outperform baseline approaches.

This letter proposes a new feature describing the trajectories of surface patches (ToSP) on human bodies for action recognition by a novel scheme of utilizing RGB and depth videos. RGB data contains appearance information by which we track specific patches on body surfaces while depth data contains spatial information by which we describe surface patches. Specifically, we use spatial-temporal interest points as initial points to track in two directions. A ToSP is extracted by keeping the neighborhood in point cloud of each point on the trajectory. By using the temporal pyramid, ToSPs are matched on several levels based on the surface feature extracted from ToSP segments. The proposed feature captures both the shape and position variations of surface patches, thus it has the advantages of trajectories and local spatial-temporal features. The experiment results show that the proposed feature outperforms the existing trajectories based features and depth features.

In many image processing and pattern recognition problems, visual contents of images are currently described by high-dimensional features, which are often redundant and noisy. Toward this end, we propose a novel unsupervised feature selection scheme, namely, nonnegative spectral analysis with constrained redundancy, by jointly leveraging nonnegative spectral clustering and redundancy analysis. The proposed method can directly identify a discriminative subset of the most useful and redundancy-constrained features. Nonnegative spectral analysis is developed to learn more accurate cluster labels of the input images, during which the feature selection is performed simultaneously. The joint learning of the cluster labels and feature selection matrix enables to select the most discriminative features. Row-wise sparse models with a general ℓ(2, p)-norm (0 < p ≤ 1) are leveraged to make the proposed model suitable for feature selection and robust to noise. Besides, the redundancy between features is explicitly exploited to control the redundancy of the selected subset. The proposed problem is formulated as an optimization problem with a well-defined objective function solved by the developed simple yet efficient iterative algorithm. Finally, we conduct extensive experiments on nine diverse image benchmarks, including face data, handwritten digit data, and object image data. The proposed method achieves encouraging the experimental results in comparison with several representative algorithms, which demonstrates the effectiveness of the proposed algorithm for unsupervised feature selection.

With the proliferation of large-scale community-contributed images, hashing-based approximate nearest neighbor search in huge databases has aroused considerable interest from the fields of computer vision and multimedia in recent years because of its computational and memory efficiency. In this paper, we propose a novel hashing method named neighborhood discriminant hashing (NDH) (for short) to implement approximate similarity search. Different from the previous work, we propose to learn a discriminant hashing function by exploiting local discriminative information, i.e., the labels of a sample can be inherited from the neighbor samples it selects. The hashing function is expected to be orthogonal to avoid redundancy in the learned hashing bits as much as possible, while an information theoretic regularization is jointly exploited using maximum entropy principle. As a consequence, the learned hashing function is compact and nonredundant among bits, while each bit is highly informative. Extensive experiments are carried out on four publicly available data sets and the comparison results demonstrate the outperforming performance of the proposed NDH method over state-of-the-art hashing techniques.

Robust principal component analysis (RPCA) is a new emerging method for exact recovery of corrupted low-rank matrices. It assumes that the real data matrix has low rank and the error matrix is sparse. This paper presents a method called double nuclear norm-based matrix decomposition (DNMD) for dealing with the image data corrupted by continuous occlusion. The method uses a unified low-rank assumption to characterize the real image data and continuous occlusion. Specifically, we assume all image vectors form a low-rank matrix, and each occlusion-induced error image is a low-rank matrix as well. Compared with RPCA, the low-rank assumption of DNMD is more intuitive for describing occlusion. Moreover, DNMD is solved by alternating direction method of multipliers. Our algorithm involves only one operator: the singular value shrinkage operator. DNMD, as a transductive method, is further extended into inductive DNMD (IDNMD). Both DNMD and IDNMD use nuclear norm for measuring the continuous occlusion-induced error, while many previous methods use L1 , L2 , or other M-estimators. Extensive experiments on removing occlusion from face images and background modeling from surveillance videos demonstrate the effectiveness of the proposed methods.

This paper aims at constructing a good graph to discover the intrinsic data structures under a semisupervised learning setting. First, we propose to build a nonnegative low-rank and sparse (referred to as NNLRS) graph for the given data representation. In particular, the weights of edges in the graph are obtained by seeking a nonnegative low-rank and sparse reconstruction coefficients matrix that represents each data sample as a linear combination of others. The so-obtained NNLRS-graph captures both the global mixture of subspaces structure (by the low-rankness) and the locally linear structure (by the sparseness) of the data, hence it is both generative and discriminative. Second, as good features are extremely important for constructing a good graph, we propose to learn the data embedding matrix and construct the graph simultaneously within one framework, which is termed as NNLRS with embedded features (referred to as NNLRS-EF). Extensive NNLRS experiments on three publicly available data sets demonstrate that the proposed method outperforms the state-of-the-art graph construction method by a large margin for both semisupervised classification and discriminative analysis, which verifies the effectiveness of our proposed method.

This article presents a simple yet effective face recognition method, called local structure-based sparse representation classification (LS_SRC). Motivated by the “divide-and-conquer” strategy, we first divide the face into local blocks and classify each local block, then integrate all the classification results to make the final decision. To classify each local block, we further divide each block into several overlapped local patches and assume that these local patches lie in a linear subspace. This subspace assumption reflects the local structure relationship of the overlapped patches, making sparse representation-based classification (SRC) feasible even when encountering the single-sample-per-person (SSPP) problem. To lighten the computing burden of LS_SRC, we further propose the local structure-based collaborative representation classification (LS_CRC). Moreover, the performance of LS_SRC and LS_CRC can be further improved by using the confusion matrix of the classifier. Experimental results on four public face databases show that our methods not only generalize well to SSPP problem but also have strong robustness to occlusion; little pose variation; and the variations of expression, illumination, and time.

The availability of massive RGB-depth (RGB-D) images poses a compelling need for effective RGB-D content understanding techniques. RGB-D images provide synchronized information from multiple views (e.g., color and depth) of real-world objects and scenes. This work proposes learning compact and discriminative features from the multiple views of RGB-D content toward effective feature representation for RGB-D image understanding. In particular, a robust multiview feature learning approach is developed, which exploits the intrinsic relations among multiple views. The feature learning in multiple views is jointly optimized in an integrated formulation. The joint optimization essentially exploits the intrinsic relations among the views, leading to effective features and making the learning process robust to noises. The feature learning function is formulated as a robust nonnegative graph embedding function over multiple graphs in various views. The graphs characterize the local geometric and discriminating structure of the multiview data. The joint sparsity in ℓ 1 -norm graph embedding and ℓ 21 -norm data factorization further enhances the robustness of feature learning. We derive an efficient computational solution for the proposed approach and provide rigorous theoretical proof with regard to its convergence. We apply the proposed approach to two RGB-D image understanding tasks: RGB-D object classification and RGB-D scene categorization. We conduct extensive experiments on two real-world RGB-D image datasets. The experimental results have demonstrated the effectiveness of the proposed approach.

Drowsy driving is one of the major causes of fatal traffic accidents. In this article, we propose a real-time system that utilizes RGB-D cameras to automatically detect driver fatigue and generate alerts to drivers. By introducing RGB-D cameras, the depth data can be obtained, which provides extra evidence to benefit the task of head detection and head pose estimation. In this system, two important visual cues (head pose and eye state) for driver fatigue detection are extracted and leveraged simultaneously. We first present a real-time 3D head pose estimation method by leveraging RGB and depth data. Then we introduce a novel method to predict eye states employing the WLBP feature, which is a powerful local image descriptor that is robust to noise and illumination variations. Finally, we integrate the results from both head pose and eye states to generate the overall conclusion. The combination and collaboration of the two types of visual cues can reduce the uncertainties and resolve the ambiguity that a single cue may induce. The experiments were performed using an inside-car environment during the day and night, and theyfully demonstrate the effectiveness and robustness of our system as well as the proposed methods of predicting head pose and eye states.

Recent years have witnessed the explosive growth of community-contributed images with rich context information, which is beneficial to the task of image retrieval. It can help us to learn a suitable metric to alleviate the semantic gap. In this paper, we propose a new distance metric learning algorithm, namely weakly-supervised deep metric learning (WDML), under the deep learning framework. It utilizes a progressive learning manner to discover knowledge by jointly exploiting the heterogeneous data structures from visual contents and user-provided tags of social images. The semantic structure in the textual space is expected to be well preserved while the problem of the noisy, incomplete or subjective tags is addressed by leveraging the visual structure in the original visual space. Besides, a sparse model with the l2,1 mixed norm is imposed on the transformation matrix of the first layer in the deep architecture to compress the noisy or redundant visual features. The proposed problem is formulated as an optimization problem with a well-defined objective function and a simple yet efficient iterative algorithm is proposed to solve it. Extensive experiments on real-world social image datasets are conducted to verify the effectiveness of the proposed method for image retrieval. Encouraging experimental results are achieved compared with several representative metric learning methods.

For the task of RGB-D object recognition, it is important to identify suitable representations of images, which can boost the performance of object recognition. In this work, we propose a novel representation learning method for RGB-D images by jointly incorporating the underlying data structure and the prior knowledge of the data. Specifically, the convolutional neural networks (CNN) are employed to learn image representation by exploiting the underlying data structure. To handle the problem of the limited RGB and depth images for object recognition, the multi-level hierarchies of features trained on ImageNet from the CNN are transferred to learn rich generic feature representation for RGB and depth images while the labeled images are leveraged. On the other hand, we propose a novel deep auto-encoders (DAE) to exploit the prior knowledge, which can overcome the expensive computational cost of optimization in feature encoding. The expected representations of images are obtained by integrating the two types of image representations. To verify the effectiveness of the proposed method, we thoroughly conduct extensive experiments on two publicly available RGB-D datasets. The encouraging experimental results compared with the state-of-the-art approaches demonstrate the advantages of the proposed method.

Question retrieval aims to increase the accessibility of the community Question Answer (cQA) archives and has attracted increasing research interests recently. In this paper, we present a novel method for improving the question retrieval performance by investigating the question term selection and weighting as well as reranking results. Different from previous work, we propose a hierarchical question classification method with a sparse regularization to mimc user's question labeling in cQAs. Based on the hierarchical classification, we explore the local context of the question for term selection and reranking results and then integrating them into our proposed general question retrieval framework. The experimental results on a Yahoo! Answers dataset show the effectiveness of our method as compared to existing general question retrieval models and some state-of-the-art methods of utilizing category information for question retrieval.

As surrogate functions of L0-norm, many nonconvex penalty functions have been proposed to enhance the sparse vector recovery. It is easy to extend these nonconvex penalty functions on singular values of a matrix to enhance low-rank matrix recovery. However, different from convex optimization, solving the nonconvex low-rank minimization problem is much more challenging than the nonconvex sparse minimization problem. We observe that all the existing nonconvex penalty functions are concave and monotonically increasing on [0, ∞). Thus their gradients are decreasing functions. Based on this property, we propose an Iteratively Reweighted Nuclear Norm (IRNN) algorithm to solve the nonconvex nonsmooth low-rank minimization problem. IRNN iteratively solves a Weighted Singular Value Thresholding (WSVT) problem. By setting the weight vector as the gradient of the concave penalty function, the WSVT problem has a closed form solution. In theory, we prove that IRNN decreases the objective function value monotonically, and any limit point is a stationary point. Extensive experiments on both synthetic data and real images demonstrate that IRNN enhances the low-rank matrix recovery compared with state-of-the-art convex algorithms.

In this paper, we propose local structure based sparse representation classification (LS SRC) to solve single sample per person (SSPP) problem. By adopting the “divide-conquer-aggregate” strategy, we successfully alleviate the dilemma of high data dimensionality and small samples, where we first divide the face into local blocks, and classify each local block, and then integrate all the classification results by voting. For each block, we further divide it into overlapped patches and assume that these patches lie in a linear subspace. This subspace assumption reflects local structure relationship of the overlapped patches and makes SRC feasible for SSPP problem. To lighten the computing burden, we further propose local structure based collaborative representation classification (LS CRC). Experimental results on three public face databases show that our methods not only generalize well to SSPP problem but also have strong robustness to expression, illumination, little pose variation, occlusion and time variation.

RGB-D object recognition is an important research topic in computer version, and seeking a robust image representation is the most important sub problem for RGB-D object recognition. On the one hand, the recently emerging deep learning methods, which learns image representations automatically by capturing the data structure, have demonstrated the impressive performance for object recognition. On the other hand, the previously commonly used hand-crafted features also encodes the prior knowledge about the data. By realizing that the hand-crafted features and machine learnt features actually characterize the different aspects of image data, rather than only using one type of feature, we propose to jointly use the machine learnt features and hand-crafted features for RGB-D object recognition. Specifically, we use the Convolution Neural Networks (CNNs) to extract the machine learnt representation, and use Locality-constrained Linear Coding (LLC) based spatial pyramid matching for hand-crafted features. We evaluated our proposed approach on three publicly available RGB-D datasets. Experimental results show that our method achieves the best performance under all the cases, which demonstrates the effectiveness of our method.

Among the flourishing amount of photos in the social media websites, "cute" images of animals are particularly attractive to the Internet users. This paper considers building an automatic model which can distinguish cute images from non-cute ones. To make the recognition results more interpretable, a lot of efforts are made to find which part of the animal appears attractive to the human users. To validate the success of our proposed method, we collect three new datasets of different animals, i.e., cats, dogs, and rabbits with both cute and non-cute images. Our model obtains promising performance in distinguishing cute images from non-cute ones. Moreover, it outperforms the classical models with not only better recognition accuracy, but also more intuitive localization of the cuteness in the images. The contribution of this paper is three-fold: (1) We collect new datasets for cuteness recognition, (2) We extend the powerful Fisher Vector representation to localize cute part in the animal recognition, and (3) Extensive experimental results show that our proposed method can recognize cute animals of cats, dogs, and rabbits.

In image retrieval, users' search intention is usually specified by textual queries, exemplar images, concept maps, and even sketches, which can only express the search intention partially. These query strategies lack the abilities to indicate the Regions Of Interests (ROIs) and represent the spatial or semantic correlations among the ROIs, which results in the so-called semantic gap between users' search intention and images' low-level visual content. In this paper, we propose a novel image search method, which allows the users to indicate any number of Regions Of Interest (ROIs) within the query as well as utilize various semantic concepts and spatial relations to search images. Specifically, we firstly propose a structured descriptor to jointly represent the categories, attributes, and spatial relations among objects. Then, based on the defined descriptor, our method ranks the images in the database according to the matching scores w.r.t. the category, attribute, and spatial relations. We conduct the experiments on the aPascal and aYahoo datasets, and experimental results show the advantage of the proposed method compared to the state of the arts.

Human action recognition from motion videos plays an important role in multimedia analysis. Different from the temporal cues of action series in motion videos, the motion tendency can also be revealed from the still images or key frames. Thus, if the action knowledge in related still images can be well adapted to the target motion videos, we would have a great chance to improve the performance of video action recognition. In this paper, we propose a framework of Still-to-Motion Adaptation (SMA) for human action recognition. Common visual features are extracted both from the related images and target videos' key frames, by which the gap between still images and videos are bridged. Meanwhile, to utilize the unlabeled training videos in target domain, we incorporate a semi-supervised process into our framework. By minimizing the difference of action prediction from still features and motion features, we formulate the still-to-motion adaptation into a joint optimization process. Experiments successfully demonstrate the effectiveness of the proposed framework and show the better performance of action recognition compared with the state-of-the-art methods. We also analyze the impact on the recognition results of target videos by knowledge adaptation from still images.

Human action recognition in videos is useful for many applications. However, there still exist huge challenges in real applications due to the variations in the appearance, lighting condition and viewing angle, of the subjects. In this consideration, depth data have advantages over red, green, blue (RGB) data because of their spatial information about the distance between object and viewpoint. Unlike existing works, we utilize the 3-D point cloud, which contains points in the 3-D real-world coordinate system to represent the external surface of human body. Specifically, we propose a new robust feature, the body surface context (BSC), by describing the distribution of relative locations of the neighbors for a reference point in the point cloud in a compact and descriptive way. The BSC encodes the cylindrical angular of the difference vector based on the characteristics of human body, which increases the descriptiveness and discriminability of the feature. As the BSC is an approximate object-centered feature, it is robust to transformations including translations and rotations, which are very common in real applications. Furthermore, we propose three schemes to represent human actions based on the new feature, including the skeleton-based scheme, the random-reference-point scheme, and the spatial-temporal scheme. In addition, to evaluate the proposed feature, we construct a human action dataset by a depth camera. Experiments on three datasets demonstrate that the proposed feature outperforms RGB-based features and other existing depth-based features, which validates that the BSC feature is promising in the field of human action recognition.

The recently proposed robust principle component analysis (RPCA) has been successfully applied in background subtraction. However, low-rank decomposition makes sense on the condition that the foreground pixels (sparsity patterns) are uniformly located at the scene, which is not realistic in real-world applications. To overcome this limitation, we reconstruct the input video frames and aim to make the foreground pixels not only sparse in space but also sparse in time. Therefore, we propose a joint video frame set division and RPCA-based method for background subtraction. In addition, we use the motion as a priori knowledge which has not been considered in the current subspace-based methods. The proposed method consists of two phases. In the first phase, we propose a lower bound-based within-class maximum division method to divide the video frame set into several subsets. In this way, the successive frames are assigned to different subsets in which the foregrounds are located at the scene randomly. In the second phase, we augment each subset using the frames with a small quantity of motion. To evaluate the proposed method, the experiments are conducted on real-world and public datasets. The comparisons with the state-of-the-art background subtraction methods validate the superiority of our method.

Distance measure between two sets of views is one central task in view-based 3D model retrieval. In this paper, we introduce a distance metric learning method for bipartite graph matching-based 3D object retrieval framework. In this method, the relationship among 3D models is formulated by a graph structure with semisupervised learning to estimate the model relevance. More specially, we model two sets of views by using a bipartite graph, on which their optimal matching is estimated. Then, we learn a refined distance metric by using the user’s relevance feedback. The proposed method has been evaluated on four data sets and the experimental results and comparison with the state-of-the-art methods demonstrate the effectiveness of the proposed method.

In recent years, sparse representation has been widely used in object recognition applications. How to learn the dictionary is a key issue to sparse representation. A popular method is to use l1 norm as the sparsity measurement of representation coefficients for dictionary learning. However, the l1 norm treats each atom in the dictionary independently, so the learned dictionary cannot well capture the multisubspaces structural information of the data. In addition, the learned subdictionary for each class usually shares some common atoms, which weakens the discriminative ability of the reconstruction error of each subdictionary. This paper presents a new dictionary learning model to improve sparse representation for image classification, which targets at learning a class-specific subdictionary for each class and a common subdictionary shared by all classes. The model is composed of a discriminative fidelity, a weighted group sparse constraint, and a subdictionary incoherence term. The discriminative fidelity encourages each class-specific subdictionary to sparsely represent the samples in the corresponding class. The weighted group sparse constraint term aims at capturing the structural information of the data. The subdictionary incoherence term is to make all subdictionaries independent as much as possible. Because the common subdictionary represents features shared by all classes, we only use the reconstruction error of each class-specific subdictionary for classification. Extensive experiments are conducted on several public image databases, and the experimental results demonstrate the power of the proposed method, compared with the state-of-the-arts.

The goal of multilabel classification is to reveal the underlying label correlations to boost the accuracy of classification tasks. Most of the existing multilabel classifiers attempt to exhaustively explore dependency between correlated labels. It increases the risk of involving unnecessary label dependencies, which are detrimental to classification performance. Actually, not all the label correlations are indispensable to multilabel model. Negligible or fragile label correlations cannot be generalized well to the testing data, especially if there exists label correlation discrepancy between training and testing sets. To minimize such negative effect in the multilabel model, we propose to learn a sparse structure of label dependency. The underlying philosophy is that as long as the multilabel dependency cannot be well explained, the principle of parsimony should be applied to the modeling process of the label correlations. The obtained sparse label dependency structure discards the outlying correlations between labels, which makes the learned model more generalizable to future samples. Experiments on real world data sets show the competitive results compared with existing algorithms.

Hashing-based similarity search techniques is becoming increasingly popular in large data sets. To capture meaningful neighbors, the topology of a data set, which represents the neighborhood relationships between its subregions and the relative proximities between the neighbors of each subregion, e.g., the relative neighborhood ranking of each subregion, should be exploited. However, most existing hashing methods are developed to preserve neighborhood relationships while ignoring the relative neighborhood proximities. Moreover, most hashing methods lack in providing a good result ranking, since there are often lots of results sharing the same Hamming distance to a query. In this paper, we propose a novel hashing method to solve these two issues jointly. The proposed method is referred to as topology preserving hashing (TPH). TPH is distinct from prior works by also preserving the neighborhood ranking. Based on this framework, we present three different TPH methods, including linear unsupervised TPH, semisupervised TPH, and kernelized TPH. Particularly, our unsupervised TPH is capable of mining semantic relationship between unlabeled data without supervised information. Extensive experiments on four large data sets demonstrate the superior performances of the proposed methods over several state-of-the-art unsupervised and semisupervised hashing techniques.

In this article, we propose a unified geolocation framework to automatically determine where on the earth a web video was shot. We analyze different social, visual, and textual relationships from a real-world dataset and find four relationships with apparent geography clues that can be used for web video geolocation. Then, the geolocation process is formulated as an optimization problem that simultaneously takes the social, visual, and textual relationships into consideration. The optimization problem is solved by an iterative procedure, which can be interpreted as a propagation of the geography information among the web video social network. Extensive experiments on a real-world dataset clearly demonstrate the effectiveness of our proposed framework, with the geolocation accuracy higher than state-of-the-art approaches.

Numerous consumer reviews of products are now available on the Internet. Consumer reviews contain rich and valuable knowledge for both firms and users. However, the reviews are often disorganized, leading to difficulties in information navigation and knowledge acquisition. This article proposes a product aspect ranking framework, which automatically identifies the important aspects of products from online consumer reviews, aiming at improving the usability of the numerous reviews. The important product aspects are identified based on two observations: 1) the important aspects are usually commented on by a large number of consumers and 2) consumer opinions on the important aspects greatly influence their overall opinions on the product. In particular, given the consumer reviews of a product, we first identify product aspects by a shallow dependency parser and determine consumer opinions on these aspects via a sentiment classifier. We then develop a probabilistic aspect ranking algorithm to infer the importance of aspects by simultaneously considering aspect frequency and the influence of consumer opinions given to each aspect over their overall opinions. The experimental results on a review corpus of 21 popular products in eight domains demonstrate the effectiveness of the proposed approach. Moreover, we apply product aspect ranking to two real-world applications, i.e., document-level sentiment classification and extractive review summarization, and achieve significant performance improvements, which demonstrate the capacity of product aspect ranking in facilitating real-world applications.

Social tagging becomes increasingly important to organize and search large-scale community-contributed photos on social websites. To facilitate generating high-quality social tags, tag recommendation by automatically assigning relevant tags to photos draws particular research interest. In this paper, we focus on the personalized tag recommendation task and try to identify user-preferred, geo-location-specific as well as semantically relevant tags for a photo by leveraging rich contexts of the freely available community-contributed photos. For users and geo-locations, we assume they have different preferred tags assigned to a photo, and propose a subspace learning method to individually uncover the both types of preferences. The goal of our work is to learn a unified subspace shared by the visual and textual domains to make visual features and textual information of photos comparable. Considering the visual feature is a lower level representation on semantics than the textual information, we adopt a progressive learning strategy by additionally introducing an intermediate subspace for the visual domain, and expect it to have consistent local structure with the textual space. Accordingly, the unified subspace is mapped from the intermediate subspace and the textual space respectively. We formulate the above learning problems into a united form, and present an iterative optimization with its convergence proof. Given an untagged photo with its geo-location to a user, the user-preferred and the geo-location-specific tags are found by the nearest neighbor search in the corresponding unified spaces. Then we combine the obtained tags and the visual appearance of the photo to discover the semantically and visually related photos, among which the most frequent tags are used as the recommended tags. Experiments on a large-scale data set collected from Flickr verify the effectivity of the proposed solution.

We modify the conventional principal component analysis (PCA) and propose a novel subspace learning framework, modified PCA (MPCA), using multiple similarity measurements. MPCA computes three similarity matrices exploiting the similarity measurements: 1) mutual information; 2) angle information; and 3) Gaussian kernel similarity. We employ the eigenvectors of similarity matrices to produce new subspaces, referred to as similarity subspaces. A new integrated similarity subspace is then generated using a novel feature selection approach. This approach needs to construct a kind of vector set, termed weak machine cell (WMC), which contains an appropriate number of the eigenvectors spanning the similarity subspaces. Combining the wrapper method and the forward selection scheme, MPCA selects a WMC at a time that has a powerful discriminative capability to classify samples. MPCA is very suitable for the application scenarios in which the number of the training samples is less than the data dimensionality. MPCA outperforms the other state-of-the-art PCA-based methods in terms of both classification accuracy and clustering result. In addition, MPCA can be applied to face image reconstruction. MPCA can use other types of similarity measurements. Extensive experiments on many popular real-world data sets, such as face databases, show that MPCA achieves desirable classification results, as well as has a powerful capability to represent data.

We present a hierarchical kernelized classification model for the automatic classification of general questions into their corresponding topic categories in community Question Answering service (cQAs). This could save many efforts of manual classification and facilitate browsing as well as better retrieving of questions from the cQA archives. To deal with the challenge of short text message of questions, we explore and optimally combine various cQA features by introducing multiple kernel learning strategy into the hierarchical classification framework. We propose a hybrid regularization approach of combining orthogonal constraint and L1 sparseness in our framework to promote the discriminative power on similar topics as well as sparsing the model parameters. The experimental results on a real world dataset from Yahoo! Answers demonstrate the effectiveness of our proposed model as compared to the state-of-the-art methods and strong baselines.

In this paper, we propose a novel Weakly-Supervised Dual Clustering (WSDC) approach for image semantic segmentation with image-level labels, i.e., collaboratively performing image segmentation and tag alignment with those regions. The proposed approach is motivated from the observation that super pixels belonging to an object class usually exist across multiple images and hence can be gathered via the idea of clustering. In WSDC, spectral clustering is adopted to cluster the super pixels obtained from a set of over-segmented images. At the same time, a linear transformation between features and labels as a kind of discriminative clustering is learned to select the discriminative features among different classes. The both clustering outputs should be consistent as much as possible. Besides, weakly-supervised constraints from image-level labels are imposed to restrict the labeling of super pixels. Finally, the non-convex and non-smooth objective function are efficiently optimized using an iterative CCCP procedure. Extensive experiments conducted on MSRC and Label Me datasets demonstrate the encouraging performance of our method in comparison with some state-of-the-arts.

Binary hashing has been widely used for efficient similarity search due to its query and storage efficiency. In most existing binary hashing methods, the high-dimensional data are embedded into Hamming space and the distance or similarity of two points are approximated by the Hamming distance between their binary codes. The Hamming distance calculation is efficient, however, in practice, there are often lots of results sharing the same Hamming distance to a query, which makes this distance measure ambiguous and poses a critical issue for similarity search where ranking is important. In this paper, we propose a weighted Hamming distance ranking algorithm (WhRank) to rank the binary codes of hashing methods. By assigning different bit-level weights to different hash bits, the returned binary codes are ranked at a finer-grained binary code level. We give an algorithm to learn the data-adaptive and query-sensitive weight for each hash bit. Evaluations on two large-scale image data sets demonstrate the efficacy of our weighted Hamming distance for binary code ranking.

In this paper, we study the robust subspace clustering problem, which aims to cluster the given possibly noisy data points into their underlying subspaces. A large pool of previous subspace clustering methods focus on the graph construction by different regularization of the representation coefficient. We instead focus on the robustness of the model to non-Gaussian noises. We propose a new robust clustering method by using the correntropy induced metric, which is robust for handling the non-Gaussian and impulsive noises. Also we further extend the method for handling the data with outlier rows/features. The multiplicative form of half-quadratic optimization is used to optimize the non-convex correntropy objective function of the proposed models. Extensive experiments on face datasets well demonstrate that the proposed methods are more robust to corruptions and occlusions.

Recent advances in Multimedia research have generated a large collection of concept models, e.g., LSCOM and Media mill 101, which become accessible to other researchers. While most current research effort still focuses on building new concepts from scratch, little effort has been made on constructing new concepts upon the existing models already in the warehouse. To address this issue, we develop a new framework in this paper, termed LEGO, to seamlessly integrate both the new target training examples and the existing primitive concept models. LEGO treats the primitive concept models as a lego toy to potentially construct an unlimited vocabulary of new concepts. Specifically, LEGO first formulates the logic operations to be the lego connectors to combine existing concept models hierarchically in probabilistic logic ontology trees. LEGO then simultaneously incorporates new target training information to efficiently disambiguate the underlying logic tree and correct the error propagation. We present extensive experimental results on a large vehicle domain data set from Image Net, and demonstrate significantly superior performance over existing state-of-the-art approaches which build new concept models from scratch.

The state-of-the-art partial-duplicate image search systems reply heavily on the match of local features like SIFT. Independently matching local features across two images ignores the overall geometry structure and therefore may incur many false matches. To reduce such matches, several geometry verification methods have been proposed. This paper introduces a new geometry verification method named as Strong Geometry Consistency (SGC), which uses the orientation, scale and location information of the local feature points to accurately and quickly remove the false matches. We also propose a simple scale weighting (SW) strategy, which gives feature points with larger scales greater weights, based on the intuition that a larger-scale feature point tends to be more robust for image search as it occupies a larger area of an image. Extensive experiments performed on three popular datasets show that SGC significantly outperforms state-of-the-art geometry verification methods, and SW can further boost the performance with marginal additional computation.

Binary hashing has been widely used for efficient similarity search. Learning efficient codes has become a research focus and it is still a challenge. In many cases, the real-world data often lies on a low-dimensional manifold, which should be taken into account to capture meaningful neighbors with hashing. The importance of a manifold is its topology, which represents the neighborhood relationships between its subregions and the relative proximities between the neighbors of each subregion, e.g. the relative ranking of neighbors of each subregion. Most existing hashing methods try to preserve the neighborhood relationships by mapping similar points to close codes, while ignoring the neighborhood rankings. Moreover, most hashing methods lack in providing a good ranking for query results since they use Hamming distance as the similarity metric, and in practice, there are often a lot of results sharing the same distance to a query. In this paper, we propose a novel hashing method to solve these two issues jointly. The proposed method is referred to as Topology Preserving Hashing (TPH). TPH is distinct from prior works by preserving the neighborhood rankings of data points in Hamming space. The learning stage of TPH is formulated as a generalized eigendecomposition problem with closed form solutions. Experimental comparisons with other state-of-the-art methods on three noted image benchmarks demonstrate the efficacy of the proposed method.

Accurately identifying the protein-ligand binding sites or pockets is of significant importance for both protein function analysis and drug design. Although much progress has been made, challenges remain, especially when the 3D structures of target proteins are not available or no homology templates can be found in the library, where the template-based methods are hard to be applied. In this paper, we report a new ligand-specific template-free predictor called TargetS for targeting protein-ligand binding sites from primary sequences. TargetS first predicts the binding residues along the sequence with ligand-specific strategy and then further identifies the binding sites from the predicted binding residues through a recursive spatial clustering algorithm. Protein evolutionary information, predicted protein secondary structure, and ligand-specific binding propensities of residues are combined to construct discriminative features; an improved AdaBoost classifier ensemble scheme based on random undersampling is proposed to deal with the serious imbalance problem between positive (binding) and negative (nonbinding) samples. Experimental results demonstrate that TargetS achieves high performances and outperforms many existing predictors. TargetS web server and data sets are freely available at: http://www.csbio.sjtu.edu.cn/bioinf/TargetS/ for academic use.

Bottom-up saliency detection aims to detect salient areas within natural images usually without learning from labeled images. Typically, the saliency map of an image is inferred by only using the information within this image (referred to as the “current image”). While efficient, such single-image-based methods may fail to obtain reliable results, because the information within a single image may be insufficient for defining saliency. In this paper, we investigate how saliency detection can benefit from the nearest neighbor structure in the image space. First, we show that existing methods can be improved by extending them to include the visual neighborhood information. This verifies the significance of the neighbors. Next, a solution of multitask sparsity pursuit is proposed to integrate the current image and its neighbors to collaboratively detect saliency. The integration is done by first representing each image as a feature matrix, and then seeking the consistently sparse elements from the joint decompositions of multiple matrices into pairs of low-rank and sparse matrices. The computational procedure is formulated as a constrained nuclear norm and ℓ2,1-norm minimization problem, which is convex and can be solved efficiently with the augmented Lagrange multiplier method. Besides the nearest neighbor structure in the visual feature space, the proposed model can also be generalized to handle multiple visual features. Extensive experiments have clearly validated its superiority over other state-of-the-art methods.

Accurate estimation of human body orientation can significantly enhance the analysis of human behavior, which is a fundamental task in the field of computer vision. However, existing orientation estimation methods cannot handle the various body poses and appearances. In this paper, we propose an innovative RGB-D-based orientation estimation method to address these challenges. By utilizing the RGB-D information, which can be real time acquired by RGB-D sensors, our method is robust to cluttered environment, illumination change and partial occlusions. Specifically, efficient static and motion cue extraction methods are proposed based on the RGB-D superpixels to reduce the noise of depth data. Since it is hard to discriminate all the 360 (°) orientation using static cues or motion cues independently, we propose to utilize a dynamic Bayesian network system (DBNS) to effectively employ the complementary nature of both static and motion cues. In order to verify our proposed method, we build a RGB-D-based human body orientation dataset that covers a wide diversity of poses and appearances. Our intensive experimental evaluations on this dataset demonstrate the effectiveness and efficiency of the proposed method.

The classical linear discriminant analysis has undergone great development and has recently been extended to different cases. In this paper, a novel discriminant subspace learning method called sparse tensor discriminant analysis (STDA) is proposed, which further extends the recently presented multilinear discriminant analysis to a sparse case. Through introducing the L1 and L2 norms into the objective function of STDA, we can obtain multiple interrelated sparse discriminant subspaces for feature extraction. As there are no closed-form solutions, k-mode optimization technique and the L1 norm sparse regression are combined to iteratively learn the optimal sparse discriminant subspace along different modes of the tensors. Moreover, each non-zero element in each subspace is selected from the most important variables/factors, and thus STDA has the potential to perform better than other discriminant subspace methods. Extensive experiments on face databases (Yale, FERET, and CMU PIE face databases) and the Weizmann action database show that the proposed STDA algorithm demonstrates the most competitive performance against the compared tensor-based methods, particularly in small sample sizes.

In this paper, we propose a novel cross-space affinity learning algorithm over different spaces with heterogeneous structures. Unlike most of affinity learning algorithms on the homogeneous space, we construct a cross-space tensor model to learn the affinity measures on heterogeneous spaces subject to a set of order constraints from the training pool. We further enhance the model with a factorization form which greatly reduces the number of parameters of the model with a controlled complexity. Moreover, from the practical perspective, we show the proposed factorized cross-space tensor model can be efficiently optimized by a series of simple quadratic optimization problems in an iterative manner. The proposed cross-space affinity learning algorithm can be applied to many real-world problems, which involve multiple heterogeneous data objects defined over different spaces. In this paper, we apply it into the recommendation system to measure the affinity between users and the product items, where a higher affinity means a higher rating of the user on the product. For an empirical evaluation, a widely used benchmark movie recommendation data set—MovieLens—is used to compare the proposed algorithm with other state-of-the-art recommendation algorithms and we show that very competitive results can be obtained.

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.

Interactive tagging is an approach that combines human and computer to assign descriptive keywords to image contents in a semi-automatic way. It can avoid the problems in automatic tagging and pure manual tagging by achieving a compromise between tagging performance and manual cost. However, conventional research efforts on interactive tagging mainly focus on sample selection and models for tag prediction. In this work, we investigate interactive tagging from a different aspect. We introduce an interactive image tagging framework that can more fully make use of human's labeling efforts. That means, it can achieve a specified tagging performance by taking less manual labeling effort or achieve better tagging performance with a specified labeling cost. In the framework, hashing is used to enable a quick clustering of image regions and a dynamic multiscale clustering labeling strategy is proposed such that users can label a large group of similar regions each time. We also employ a tag refinement method such that several inappropriate tags can be automatically corrected. Experiments on a large dataset demonstrate the effectiveness of our approach

Spatial contexts encode rich discriminative information for visual classification. However, as object shapes and scales vary significantly among images, spatial contexts with manually specified distance ranges are not guaranteed with optimality. In this work, we investigate how to automatically select discriminative and stable distance bin groups for modeling image spatial contexts to improve classification performance. We make two observations. First, the number of distance bins for context modeling can be arbitrarily large, and discriminative contexts are only from a small subset of distance bins. Second, adjacent distance bins for contexts modeling often show similar characteristics, thus encouraging grouping them together can result in more stable representation. Utilizing these two observations, we propose an omni-range spatial context mining framework for image classification. A sparse selection and grouping regularizer is employed along with an empirical risk, to discover discriminative and stable distance bin groups for context modeling. To facilitate efficient optimization, the objective function is approximated by a smooth convex function with theoretically guaranteed error bounds. The selected and grouped image spatial contexts, which are applied in food and national flag recognition, are demonstrated to be discriminative, compact and robust.

The use of local features for image representation has been proven very effective for a variety of visual tasks such as object localization and scene classification. However, local image features carry little semantic information which is potentially not enough for high level visual tasks. To solve this problem, in this paper, we propose to use a supervised semantic image representation for scene classification, where an image is represented as a response histogram. This response histogram is a combination of the prediction of pre-trained generic object classifiers and classifiers generated by supervised learning. Besides, the use of sparsity constraints makes the proposed representation more efficient and effective to compute. Performances on the UIUC-Sports dataset, the MIT Indoor scene dataset and the Scene-15 dataset demonstrate the effectiveness of the proposed method.

With the popularity of social media applications, large amounts of social images associated with rich context are available, which is helpful for many applications. In this paper, we propose a Low Rank distance Metric Learning (LRML) algorithm by discovering knowledge from these rich contextual data, to boost the performance of CBIR. Different from traditional approaches that often use the must-links and cannot-links between images, the proposed method exploits information from the visual and textual domains. We assume that the visual similarity estimated by the learned metric is expected to be consistent with the semantic similarity in the textual domain. Since tags are usually noisy, misspelling or meaningless, we also leverage the preservation of visual structure to prevent overfitting those noisy tags. On the other hand, the metric is straightforward constrained to be low rank. We formulate it as a convex optimization problem with nuclear norm minimization and propose an effective optimization algorithm based on proximal gradient method. With the learned metric for image retrieval, some experimental evaluations on a real-world dataset demonstrate the outperformance of our approach over other related work.

Query expansion has been successfully employed to improve the performance of image retrieval system. It usually expands the original query based on the information from top ranked images. However, it may fail when some of the top ranked images are false positive or contain noisy features. To minimize the amount of irrelevant local features introduced, we propose to enhance query expansion by fast binary matching. More specifically, the noisy points on a candidate image are filtered out by local verification with their mapped locations on the query image. We further rank the expansion results by three different measurements based on local patch similarity in the image space. Experiments on partial-duplicate Web image search with a database of one million images show that the proposed approach achieves promising improvement in mean Average Precision (mAP) over the state-of-the-art query expansion approaches, and remains efficient in search time.

In this demo, we present an effective real-time system for ATM intelligent monitoring by using Kinect of Microsoft. With Kinect, we can easily detect people in ATM room and get their position information. By analyzing position information and video content, the system detects abnormal behaviors such as face-hiding, peeping and wandering, while records the time of these abnormal videos. Therefore, it not only prevents crimes but also helps to find suspects quickly after crimes have happened. The experimental results show that the system has the advantages of robustness, and provides a new mean for preventing financial crimes.

Recently, extensive research efforts have been dedicated to view-based methods for 3-D object retrieval due to the highly discriminative property of multiviews for 3-D object representation. However, most of state-of-the-art approaches highly depend on their own camera array settings for capturing views of 3-D objects. In order to move toward a general framework for 3-D object retrieval without the limitation of camera array restriction, a camera constraint-free view-based (CCFV) 3-D object retrieval algorithm is proposed in this paper. In this framework, each object is represented by a free set of views, which means that these views can be captured from any direction without camera constraint. For each query object, we first cluster all query views to generate the view clusters, which are then used to build the query models. For a more accurate 3-D object comparison, a positive matching model and a negative matching model are individually trained using positive and negative matched samples, respectively. The CCFV model is generated on the basis of the query Gaussian models by combining the positive matching model and the negative matching model. The CCFV removes the constraint of static camera array settings for view capturing and can be applied to any view-based 3-D object database. We conduct experiments on the National Taiwan University 3-D model database and the ETH 3-D object database. Experimental results show that the proposed scheme can achieve better performance than state-of-the-art methods.

User interaction is an effective way to handle the semantic gap problem in image annotation. To minimize user effort in the interactions, many active learning methods were proposed. These methods treat the semantic concepts individually or correlatively. However, they still neglect the key motivation of user feedback: to tackle the semantic gap. The size of the semantic gap of each concept is an important factor that affects the performance of user feedback. User should pay more efforts to the concepts with large semantic gaps, and vice versa. In this paper, we propose a semantic-gap-oriented active learning method, which incorporates the semantic gap measure into the information-minimization-based sample selection strategy. The basic learning model used in the active learning framework is an extended multilabel version of the sparse-graph-based semisupervised learning method that incorporates the semantic correlation. Extensive experiments conducted on two benchmark image data sets demonstrated the importance of bringing the semantic gap measure into the active learning process.

No abstract available.

The thirteen papers in this special issue focus on effective techniques for learning semantics from multimedia Web resources.

In this work, we investigate how to reassign the fully annotated labels at image level to those contextually derived semantic regions, namely Label-to-Region (L2R), in a collective manner. Given a set of input images with label annotations, the basic idea of our approach to L2R is to first discover the patch correspondence across images, and then propagate the common labels shared in image pairs to these correlated patches. Specially, our approach consists of following aspects. First, each of the input images is encoded as a Bag-of-Hierarchical-Patch (BOP) for capturing the rich cues at variant scales, and the individual patches are expressed by patch-level feature descriptors. Second, we present a sparse representation formulation for discovering how well an image or a semantic region can be robustly reconstructed by all the other image patches from the input image set. The underlying philosophy of our formulation is that an image region can be sparsely reconstructed with the image patches belonging to the other images with common labels, while the robustness in label propagation across images requires that these selected patches come from very few images. This preference of being sparse at both patch and image level is namedbi-layer sparsity prior. Meanwhile, we enforce the preference of choosing larger-size patches in reconstruction, referred to ascontinuity-biased priorin this work, which may further enhance the reliability of L2R assignment. Finally, we harness the reconstruction coefficients to propagate the image labels to the matched patches, and fuse the propagation results over all patches to finalize the L2R task. As a by-product, the proposed continuity-biased bi-layer sparse representation formulation can be naturally applied to perform image annotation on new testing images. Extensive experiments on three public image datasets clearly demonstrate the effectiveness of our proposed framework in both L2R assignment and image annotation.

To bridge the semantic gap between low level feature and human perception, most of the existing algorithms aim mainly at annotating images with concepts coming from only one semantic space, e.g. cognitive or affective. The naive combination of the outputs from these spaces will implicitly force the conditional independence and ignore the correlations among the spaces. In this paper, to exploit the comprehensive semantic of images, we propose a general framework for harmoniously integrating the above multiple semantics, and investigating the problem of learning to annotate images with training images labeled in two or more correlated semantic spaces, such as fascinating nighttime, or exciting cat. This kind of semantic annotation is more oriented to real world search scenario. Our proposed approach outperforms the baseline algorithms by making the following contributions. 1) Unlike previous methods that annotate images within only one semantic space, our proposed multi-semantic annotation associates each image with labels from multiple semantic spaces. 2) We develop a multi-task linear discriminative model to learn a linear mapping from features to labels. The tasks are correlated by imposing the exclusive group lasso regularization for competitive feature selection, and the graph Laplacian regularization to deal with insufficient training sample issue. 3) A Nesterov-type smoothing approximation algorithm is presented for efficient optimization of our model. Extensive experiments on NUS-WIDEEmotive dataset (56k images) with 8×81 emotive cognitive concepts and Object&Scene datasets from NUS-WIDE well validate the effectiveness of the proposed approach.

We present a novel system, named PhotoReference, to help users, especially amateur photographers, to interactively learn to improve photograph skills by leveraging the available web image collections. The idea behind is based on the observation that the quality of a photo is mainly determined with the important camera parameters set by users when the photo is taken. In the proposed system, given a user-provided photo and its associated metadata, the system first retrieves sets of high-quality community-contributed photos with similar visual content and shoot settings to the input photo. Then users are allowed to interactively explore the returned photos as per different camera parameters, such that they can easily share and learn the photograph experiences of other people; and more importantly, users are able to get intuitive knowledge on the relationship between camera parameters and the resulted shooting effects and thus can use it to guide their future photograph activities.

In this article, we exploit the problem of annotating a large-scale image corpus by label propagation over noisily tagged web images. To annotate the images more accurately, we propose a novel k NN-sparse graph-based semi-supervised learning approach for harnessing the labeled and unlabeled data simultaneously. The sparse graph constructed by datum-wise one-vs- k NN sparse reconstructions of all samples can remove most of the semantically unrelated links among the data, and thus it is more robust and discriminative than the conventional graphs. Meanwhile, we apply the approximate k nearest neighbors to accelerate the sparse graph construction without loosing its effectiveness. More importantly, we propose an effective training label refinement strategy within this graph-based learning framework to handle the noise in the training labels, by bringing in a dual regularization for both the quantity and sparsity of the noise. We conduct extensive experiments on a real-world image database consisting of 55,615 Flickr images and noisily tagged training labels. The results demonstrate both the effectiveness and efficiency of the proposed approach and its capability to deal with the noise in the training labels.

The explosive growth of Web videos brings out the challenge of how to efficiently browse hundreds or even thousands of videos at a glance. Given an event-driven query, social media Web sites usually return a large number of videos that are diverse and noisy in a ranking list. Exploring such results will be time-consuming and thus degrades user experience. This article presents a novel scheme that is able to summarize the content of video search results by mining and threading “key” shots, such that users can get an overview of main content of these videos at a glance. The proposed framework mainly comprises four stages. First, given an event query, a set of Web videos is collected associated with their ranking order and tags. Second, key-shots are established and ranked based on near-duplicate keyframe detection and they are threaded in a chronological order. Third, we analyze the tags associated with key-shots. Irrelevant tags are filtered out via a representativeness and descriptiveness analysis, whereas the remaining tags are propagated among key-shots by random walk. Finally, summarization is formulated as an optimization framework that compromises relevance of key-shots and user-defined skimming ratio. We provide two types of summarization: video skimming and visual-textual storyboard. We conduct user studies on twenty event queries for over hundred hours of videos crawled from YouTube. The evaluation demonstrates the feasibility and effectiveness of the proposed solution.

In this work, we investigate how to propagate annotated labels for a given single image from the image-level to their corresponding semantic regions, namely Label-to-Region (L2R), by utilizing the auxiliary knowledge from Internet image search with the annotated image labels as queries. A nonparametric solution is proposed to perform L2R for single image with complete labels. First, each label of the image is used as query for online image search engines to obtain a set of semantically related and visually similar images, which along with the input image are encoded as Bags-of-Hierarchical-Patches. Then, an efficient two-stage feature mining procedure is presented to discover those input-image specific, salient and descriptive features for each label from the proposed Interpolation SIFT (iSIFT) feature pool. These features consequently constitute a patch-level representation, and the continuity-biased sparse coding is proposed to select few patches from the online images with preference to larger patches to reconstruct a candidate region, which randomly merges the spatially connected patches of the input image. Such candidate regions are further ranked according to the reconstruction errors, and the top regions are used to derive the label confidence vector for each patch of the input image. Finally, a patch clustering procedure is performed as postprocessing to finalize L2R for the input image. Extensive experiments on three public databases demonstrate the encouraging performance of the proposed nonparametric L2R solution.

In this paper, we present a travel guidance system W2Go (Where to Go), which can automatically recognize and rank the landmarks for travellers. In this system, a novel Automatic Landmark Ranking (ALR) method is proposed by utilizing the tag and geo-tag information of photos in Flickr and user knowledge from Yahoo Travel Guide. ALR selects the popular tourist attractions (landmarks) based on not only the subjective opinion of the travel editors as is currently done on sites like WikiTravel and Yahoo Travel Guide, but also the ranking derived from popularity among tourists. Our approach utilizes geo-tag information to locate the positions of the tag-indicated places, and computes the probability of a tag being a landmark/site name. For potential landmarks, impact factors are calculated from the frequency of tags, user numbers in Flickr, and user knowledge in Yahoo Travel Guide. These tags are then ranked based on the impact factors. Several representative views for popular landmarks are generated from the crawled images with geo-tags to describe and present them in context of information derived from several relevant reference sources. The experimental comparisons to the other systems are conducted on eight famous cities over the world. User-based evaluation demonstrates the effectiveness of the proposed ALR method and the W2Go system.

Targeting the same objective of alleviating the manual work as automatic annotation, in this paper, we propose a novel framework with minimal human effort to manually annotate a large-scale image corpus. In this framework, a dynamic multi-scale cluster labeling strategy is proposed to manually label the clusters of similar image regions. The users label the multi-scale clusters of regions instead of individual images, thus each labeling operation can annotate hundreds or even thousands of images simultaneously with much reduced manual work. Meanwhile the manual labeling guarantees the accuracy of the labels. Compared to automatic annotation, the proposed framework is more flexible, general and effective, especially for annotating those labels with large semantic gaps. Experiments on NUS-WIDE dataset demonstrate that the proposed fast manual annotation framework is much more effective than automatic annotation and comparatively efficient.

This paper presents a system for automatically detecting and analyzing complex player actions in moving background sports video sequences, aiming at action-based sports videos indexing and providing kinematic measurements for coach assistance and performance improvement. The system works in a coarse-to-fine fashion. For an input video, in the coarse granularity level, we automatically segment the highlights, that is, the video clips containing the desired action as summaries for general user viewing purposes; in the middle granularity level, we recognize the action types to support action-based video indexing and retrieval; and finally in the fine granularity level, the critical kinematic parameters of player action are obtained for sports professionals' training purposes. However, the complex and dynamic background of sports videos and the complexity of player actions bring considerable difficulty to the automatic analysis. To fulfill such a challenging task, robust algorithms including global motion estimation with adaptive outliers filtering, object segmentation based on adaptive background construction, and automatic human body tracking are proposed in this paper. Two visual analyzing tools: motion panorama and overlay composition, are also introduced. Real diving and jump game videos are used to test the proposed system and algorithms, and the extensive and encouraging experimental results show their effectiveness.

The goal of image classification is to classify a collection of unlabeled images into a set of semantic classes. Many methods have been proposed to approach this goal by leveraging visual appearances of local patches in images. However, the spatial context between these local patches also provides significant information to improve the classification accuracy. Traditional spatial contextual models, such as two-dimensional hidden Markov model, attempt to construct one common model for each image category to depict the spatial structures of the images in this class. However due to large intra-class variances in an image category, one single model has difficulties in representing various spatial contexts in different images. In contrast, we propose to construct a prototype set of spatial contextual models by leveraging the kernel methods rather than only one model. Such an algorithm combines the advantages of rich representation ability of spatial contextual models as well as the powerful classification ability of kernel method. In particular, we propose a new distance measure between different spatial contextual models by integrating joint appearance-spatial image features. Such a distance measure can be efficiently computed in a recursive formulation that scales well to image size. Extensive experiments demonstrate that the proposed approach significantly outperforms the state-of-the-art approaches.

In most of the learning-based image annotation approaches, images are represented using multiple-instance (local) or single-instance (global) features. Their performances, however, are mixed as for certain concepts, the single-instance representations of images are more suitable, while for others, the multiple-instance representations are better. Thus this paper explores a unified learning framework that combines the multiple-instance and single-instance representations for image annotation. More specifically, we propose an integrated graph-based semi-supervised learning framework to utilize these two types of representations simultaneously. We further explore three strategies to convert from multiple-instance representation into a single-instance one. Experiments conducted on the COREL image dataset demonstrate the effectiveness and efficiency of the proposed integrated framework and the conversion strategies.

This paper proposes an efficient sparse metric learning algorithm in high dimensional space via an l1-penalized log-determinant regularization. Compare to the most existing distance metric learning algorithms, the proposed algorithm exploits the sparsity nature underlying the intrinsic high dimensional feature space. This sparsity prior of learning distance metric serves to regularize the complexity of the distance model especially in the "less example number p and high dimension d" setting. Theoretically, by analogy to the covariance estimation problem, we find the proposed distance learning algorithm has a consistent result at rate O (√m2 log d)/n) to the target distance matrix with at most m nonzeros per row. Moreover, from the implementation perspective, this l1-penalized log-determinant formulation can be efficiently optimized in a block coordinate descent fashion which is much faster than the standard semi-definite programming which has been widely adopted in many other advanced distance learning algorithms. We compare this algorithm with other state-of-the-art ones on various datasets and competitive results are obtained.

10.1145/1631272.1631551

In this work, we investigate how to automatically reassign the manually annotated labels at the image-level to those contextually derived semantic regions. First, we propose a bi-layer sparse coding formulation for uncovering how an image or semantic region can be robustly reconstructed from the over-segmented image patches of an image set. We then harness it for the automatic label to region assignment of the entire image set. The solution to bi-layer sparse coding is achieved by convex l1-norm minimization. The underlying philosophy of bi-layer sparse coding is that an image or semantic region can be sparsely reconstructed via the atomic image patches belonging to the images with common labels, while the robustness in label propagation requires that these selected atomic patches come from very few images. Each layer of sparse coding produces the image label assignment to those selected atomic patches and merged candidate regions based on the shared image labels. The results from all bi-layer sparse codings over all candidate regions are then fused to obtain the entire label to region assignments. Besides, the presenting bi-layer sparse coding framework can be naturally applied to perform image annotation on new test images. Extensive experiments on three public image datasets clearly demonstrate the effectiveness of our proposed framework in both label to region assignment and image annotation tasks.

This paper presents a novel system to explore places of interests based on the large amount of photos that are placed on Google Maps. The system, named ViewFocus, estimates the view directions of photos via robust object matching and camera reconstruction techniques, and geo-registers the directions on the map. Thus users are able to select the places they are interested in, and the system automatically returns a set of precise photos of the target places for users to focus their exploration, by filtering out photos that are pointing to other directions.

We present PornProbe, a pornography detection system that detects pornographic contents in videos. To build such a detection system, we leverage a large scale training data set with 65,827 positive training image samples out of a total of 420,615 training samples, and a novel detection scheme based on hierarchical LDA-SVM. The system combines the unsupervised clustering in Latent Dirichlet Allocation (LDA) and supervised learning in Support Vector Machine, so as to achieve both high precision and recall while ensuring efficiency in both training and testing. This demonstration shows how the system detects the pornographic scenes in restricted artistic (RA) movies.

10.1145/1631272.1631305

Image fusion has been over-studied recently. Nevertheless, few works aim to how to evaluate the performance of image fusion algorithms. In this paper, we extend the work in image quality evaluation to a novel metric for objective evaluation of image fusion. Firstly the input images and the result image are converted into local sensitive intensity (LSI) by Radon transform. Then we use the sensitive intensity to measure how many information have been transferred from each source into the fused result by the difference of LSI. Finally all the LSI pairs are incorporated into the expression according to Weber-Fechner law. Experimental results demonstrate that our proposed metric is compliant with subjective evaluations and outperforms other recently developed objective metrics of image fusion.

Conventional active learning dynamically constructs the training set only along the sample dimension. While this is the right strategy in binary classification, it is suboptimal for multilabel image classification. We argue that for each selected sample, only some effective labels need to be annotated while others can be inferred by exploring the label correlations. The reason is that the contributions of different labels to minimizing the classification error are different due to the inherent label correlations. To this end, we propose to select sample-label pairs, rather than only samples, to minimize a multilabel Bayesian classification error bound. We call it two-dimensional active learning because it considers both the sample dimension and the label dimension. Furthermore, as the number of training samples increases rapidly over time due to active learning, it becomes intractable for the offline learner to retrain a new model on the whole training set. So we develop an efficient online learner to adapt the existing model with the new one by minimizing their model distance under a set of multilabel constraints. The effectiveness and efficiency of the proposed method are evaluated on two benchmark data sets and a realistic image collection from a real-world image sharing Web site-Corbis.

Learning-based video annotation is a promising approach to facilitating video retrieval and it can avoid the intensive labor costs of pure manual annotation. But it frequently encounters several difficulties, such as insufficiency of training data and the curse of dimensionality. In this paper, we propose a method named optimized multigraph-based semi-supervised learning (OMG-SSL), which aims to simultaneously tackle these difficulties in a unified scheme. We show that various crucial factors in video annotation, including multiple modalities, multiple distance functions, and temporal consistency, all correspond to different relationships among video units, and hence they can be represented by different graphs. Therefore, these factors can be simultaneously dealt with by learning with multiple graphs, namely, the proposed OMG-SSL approach. Different from the existing graph-based semi-supervised learning methods that only utilize one graph, OMG-SSL integrates multiple graphs into a regularization framework in order to sufficiently explore their complementation. We show that this scheme is equivalent to first fusing multiple graphs and then conducting semi-supervised learning on the fused graph. Through an optimization approach, it is able to assign suitable weights to the graphs. Furthermore, we show that the proposed method can be implemented through a computationally efficient iterative process. Extensive experiments on the TREC video retrieval evaluation (TRECVID) benchmark have demonstrated the effectiveness and efficiency of our proposed approach.

In the past few years, video annotation has benefited a lot from the progress of machine learning techniques. Recently, graph-based semi-supervised learning has gained much attention in this domain. However, as a crucial factor of these algorithms, the estimation of pairwise similarity has not been sufficiently studied. Generally, the similarity of two samples is estimated based on the Euclidean distance between them. But we will show that the similarity between two samples is not merely related to their distance but also related to the distribution of surrounding samples and labels. It is shown that the traditional distance-based similarity measure may lead to high classification error rates even on several simple datasets. To address this issue, we propose a novel neighborhood similarity measure, which explores the local sample and label distributions. We show that the neighborhood similarity between two samples simultaneously takes into account three characteristics: 1) their distance; 2) the distribution difference of the surrounding samples; and 3) the distribution difference of surrounding labels. Extensive experiments have demonstrated the superiority of neighborhood similarity over the existing distance-based similarity.

Recently, graph-based semisupervised learning methods have been widely applied in multimedia research area. However, for the application of video semantic annotation in multilabel setting, these methods neglect an important characteristic of video data: The semantic concepts appear correlatively and interact naturally with each other rather than exist in isolation. In this paper, we adapt this semantic correlation into graph-based semisupervised learning and propose a novel method named correlative linear neighborhood propagation to improve annotation performance. Experiments conducted on the Text REtrieval Conference VIDeo retrieval evaluation data set have demonstrated its effectiveness and efficiency.

In this paper, we propose a two-dimensional active learning scheme and show its application in image classification. Traditional active learning methods select samples only along the sample dimension. While this is the right strategy in binary classification, it is sub-optimal for multi-label classification. In multi-label classification, we argue that, for each selected sample, only a part of more effective labels are necessary to be annotated while others can be inferred by exploring the correlations among the labels. The reason is that the contributions of different labels to minimizing the classification error are different due to the inherent label correlations. To this end, we propose to select sample-label pairs, rather than only samples, to minimize a multi-label Bayesian classification error bound. This new active learning strategy not only considers the sample dimension but also the label dimension, and we call it Two-Dimensional Active Learning (2DAL). We also show that the traditional active learning formulation is a special case of 2DAL when there is only one label. Extensive experiments conducted on two real-world applications show that the 2DAL significantly outperforms the best existing approaches which did not take label correlation into account.

The goal of image categorization is to classify a collection of unlabeled images into a set of predefined classes to support semantic-level image retrieval. The distance measures used in most existing approaches either ignored the spatial structures or used them in a separate step. As a result, these distance measures achieved only limited success. To address these difficulties, in this paper, we propose a new distance measure that integrates joint appearance-spatial image features. Such a distance measure is computed as an upper bound of an information-theoretic discrimination, and can be computed efficiently in a recursive formulation that scales well to image size. In addition, the upper bound approximation can be further tightened via adaption learning from a universal reference model. Extensive experiments on two widely-used data sets show that the proposed approach significantly outperforms the state-of-the-art approaches.

Conventional approaches to video search reranking only care whether search results are relevant or irrelevant to the given query, while the ranking order of these results indicating the level of relevance or typicality are usually neglected. This paper presents a novel learning-based approach to video search reranking by investigating the ranking order information. The proposed approach, called pseudo preference feedback (PPF), automatically discovers an optimal set of pseudo preference pairs from the initial ranked list and learns a reranking model by ranking support vector machines (ranking SVM) based on the selected pairs. We have proved that PPF can be used for any reranking purpose such as video search and concept detection. We conducted comprehensive experiments for both automatic search and concept detection tasks over TRECVID 2006-2007 benchmark, and showed that PPF could gain significant improvements over the baselines.

10.1145/1459359.1459494

Recently, many learning methods based on multiple-instance (local) or single-instance (global) representations of images have been proposed for image annotation. Their performances on image annotation, however, are mixed as for certain concepts the single-instance representations of images are more suitable, while for some other concepts the multiple-instance representations are better. Thus in this paper, we explore an unified learning framework that combines the multiple-instance and single-instance representations for image annotation. More specifically, we propose an integrated graph-based semi-supervised learning framework to utilize these two types of representations simultaneously, and explore an effective and computationally efficient strategy to convert the multiple-instance representation into a single-instance one. Experiments conducted on the Coral image dataset show the effectiveness and efficiency of the proposed integrated framework.

This paper presents a novel learning-based method, called “multi-layer multi-instance (MLMI) learning,” for video concept detection. Most of existing methods have treated video as a flat data sequence and have not investigated theintrinsic hierarchy structureof the video content deeply. However, video is essentially a kind of media with ML structure. For example, a video can be represented by a hierarchical structure including, from large to small,shot,frame, andregion, where each pair of contiguous layers fits the typical MI setting. We call such a ML structure and the MI relations embedded in the structure as the MLMI setting. In this paper, we systematically study both ML structure and MI relations embedded in video content by formulating video concept detection as a MLMI learning problem. Specifically, we first construct a MLMI kernel to simultaneously model such ML structure and MI relations. To deal with theambiguity propagationproblem which is introduced by weak labeling and ML structure, we then propose a regularization framework which takeshyper-bagprediction error, sublayer prediction error, inter-layer inconsistency measure, and classifier complexity into consideration. We have applied the proposed MLMI learning method to concept detection task over TRECVid 2005 development corpus, and report better performance to vector-based and the state-of-the-art MI learning methods.

The insufficiency of labeled training data for representing the distribution of the entire dataset is a major obstacle in automatic semantic annotation of large-scale video database. Semi-supervised learning algorithms, which attempt to learn from both labeled and unlabeled data, are promising to solve this problem. In this paper, a novel graph-based semi-supervised learning method namedkernellinearneighborhoodpropagation(KLNP) is proposed and applied to video annotation. This approach combines theconsistencyassumption, which is the basic assumption in semi-supervised learning, and thelocallinearembedding(LLE) method in a nonlinear kernel-mapped space. KLNP improves a recently proposed methodlinearneighborhoodpropagation(LNP) by tackling the limitation of its local linear assumption on the distribution of semantics. Experiments conducted on the TRECVID data set demonstrate that this approach outperforms other popular graph-based semi-supervised learning methods for video semantic annotation.

Automatic video annotation is an important ingredient for semantic-level video browsing, search and navigation. Much attention has been paid to this topic in recent years. These researches have evolved through two paradigms. In the first paradigm, each concept is individually annotated by a pre-trained binary classifier. However, this method ignores the rich information between the video concepts and only achieves limited success. Evolved from the first paradigm, the methods in the second paradigm add an extra step on the top of the first individual classifiers to fuse the multiple detections of the concepts. However, the performance of these methods can be degraded by the error propagation incurred in the first step to the second fusion one. In this article, another paradigm of the video annotation method is proposed to address these problems. It simultaneously annotates the concepts as well as model correlations between them in one step by the proposed Correlative Multilabel (CML) method, which benefits from the compensation of complementary information between different labels. Furthermore, since the video clips are composed by temporally ordered frame sequences, we extend the proposed method to exploit the rich temporal information in the videos. Specifically, a temporal-kernel is incorporated into the CML method based on the discriminative information between Hidden Markov Models (HMMs) that are learned from the videos. We compare the performance between the proposed approach and the state-of-the-art approaches in the first and second paradigms on the widely used TRECVID data set. As to be shown, superior performance of the proposed method is gained.

We propose a new multiple instance learning (MIL) algorithm to learn image categories. Unlike existing MIL algorithms, in which the individual instances in a bag are assumed to be independent with each other, we develop concurrent tensors to explicitly model the inter-dependency between the instances to better capture image's inherent semantics. Rank-1 tensor factorization is then applied to obtain the label of each instance. Furthermore, we formulate the classification problem in the reproducing kernel Hilbert space (RKHS) to extend instance label prediction to the whole feature space. Finally, a regularizer is introduced, which avoids overfitting and significantly improves learning machine's generalization capability, similar to that in SVMs. We report superior categorization performances compared with key existing approaches on both the COREL and the Caltech datasets.

As a major family of semi-supervised learning, graph based semi-supervised learning methods have attracted lots of interests in the machine learning community as well as many application areas recently. However, for the application of video semantic annotation, these methods only consider the relations among samples in the feature space and neglect an intrinsic property of video data: the temporally adjacent video segments (e.g., shots) usually have similar semantic concept. In this paper, we adapt this temporal consistency property of video data into graph based semi-supervised learning and propose a novel method named temporally consistent Gaussian random field (TCGRF) to improve the annotation results. Experiments conducted on the TREC VID data set have demonstrated its effectiveness.

In this paper, we present a novel framework for video semantic detection based on transductive inference and hierarchical clustering, which directly focuses on predicting the available samples in a current unlabeled pool, instead of trying to build a classifier workable for any unavailable data. In this framework, a number of hierarchical clustering results are constructed from theentire video datasetcontaining both labeled and unlabeled examples. We aim to make the clusters aspureas possible, i.e., samples in a same cluster mostly have a same label. To furtherpurifythese hierarchical clustering results, an EM based cluster-tuning algorithm is iteratively employed. Based on these clustering results, several hypotheses are generated byprobability votingamong labeled samples in the obtained clusters. From these hypotheses, one of them is chosen according to theVapnik combined bound, and it is then applied to predict the labels of unlabeled samples. This selected transductive hypothesis, which is only interested in predicting the available unlabeled samples in test set rather than producing a general classifier like inductive inference learning, exploits the structure and distribution of the unlabeled pool to achieve a minimal test error bound. Thus it can have better generalization ability for video annotation both theoretically and experimentally. This is also shown by our experiment results.

In this paper, we address the issue of typicality ranking for video annotation and propose to use a novel criterion, average typicality precision (ATP), to replace the frequently used one, average precision (AP), for evaluating the performance of video annotation algorithms. General annotation methods just care the number of true-positive samples at the top of the ranked list; they actually do not care the order of these samples. We argue that it is more reasonable to rank "typical" true-positive samples higher than non-typical ones, which can be evaluated by our proposed ATP. However, generally the labels of the training data only differentiate true from false; that is to say, typical or non-typical training samples have the same contribution to the learning process. Therefore, the labels of the unlabeled data learned from these training data can not well measure the typicality. In this paper, we relax the labels of the training data to real-valued typicality scores by a pre-processing stage, which is accomplished by three approaches, including density estimation, user feedback and active learning. Then the typicality scores of the training data are propagated to unlabeled data using manifold-ranking. Experiments conducted on the TRECVID data set demonstrate that this typicality ranking scheme is more consistent with human perception than normal accuracy based ranking schemes.

Graph-based semi-supervised learning (SSL) has attracted lots of interests in machine learning community as well as many application areas including video annotation recently. However, one of the two basic assumptions, structure assumption, which is an essential point of graph-based SSL, is not embedded into the pairwise similarity measure. Accordingly, we propose a novel graph-based SSL method for video annotation, named anisotropic manifold ranking (AniMR), based on a structure-related similarity measure. This method takes the influence of the density difference between samples into account to improve the pairwise similarity. Furthermore, we will show that AniMR can also be deduced from partial differential equation (PDE) based anisotropic diffusion. It demonstrates that the label propagation in AniMR is anisotropic, which is intrinsically different from the isotropic label propagation process in general graph-based SSL methods. Experiments conducted on the TRECVID data set show this approach outperforms ordinary graph-based SSL methods and is effective for video semantic annotation.

In video concept detection, most existing methods have not well studied the intrinsic hierarchical structure of video content. However, unlike flat attribute-value data used in many existing methods, video is essentially a structured media with multi-layer representation. For example, a video can be represented by a hierarchical structure including, from large to small, shot, key-frame, and region. Moreover, it fits the typical Multi-Instance (MI) setting in which the "bag-instance" correspondence is embedded among contiguous layers. We call such multi-layer structure and the "bag-instance" relation embedded in the structure as Multi-Layer Multi-Instance (MLMI) setting in this paper. We formulate video concept detection as an MLMI learning problem in which a rooted tree with MLMI nature embedded is devised to represent a video segment. Furthermore, by fusing the information from different layers, we construct a novel MLMI kernel to measure the similarities between the instances in the same and different layers. In contrast to traditional MI learning, both the Multi-Layer structure and Multi-Instance relations are leveraged simultaneously in the proposed kernel. We applied MLMI kernel to concept detection task on TRECVID 2005 corpus and reported superior performance (+25% in Mean Average Precision) to standard Support Vector Machine based approaches.

Automatically annotating concepts for video is a key to semantic-level video browsing, search and navigation. The research on this topic evolved through two paradigms. The first paradigm used binary classification to detect each individual concept in a concept set. It achieved only limited success, as it did not model the inherent correlation between concepts, e.g., urban and building. The second paradigm added a second step on top of the individual concept detectors to fuse multiple concepts. However, its performance varies because the errors incurred in the first detection step can propagate to the second fusion step and therefore degrade the overall performance. To address the above issues, we propose a third paradigm which simultaneously classifies concepts and models correlations between them in a single step by using a novel Correlative Multi-Label (CML) framework. We compare the performance between our proposed approach and the state-of-the-art approaches in the first and second paradigms on the widely used TRECVID data set. We report superior performance from the proposed approach.

Most of the existing methods for natural scene categorization only consider whether a sample is relevant or irrelevant to a particular concept. However, for the samples relevant to a certain concept, their typicalities or relevancy scores to the concept generally are different. Typicality measure should be taken into account to make the categorization results more consistent with human's perception. In this paper, we propose a novel typicality ranking scheme for categorizing natural scenes through a two-stage semi-supervised multiple-instance learning method. The first stage infers the typicalities of the underlying positive instances (i.e., regions in images) in the training dataset and the second one predicts the typicality of each bag (i.e., image) in a semi-supervised manner. Compared to existing typicality ranking approaches, the main advantages of the proposed method lie in twofold. First, it only needs image-level labels instead of region-level ones in the training stage. Second, it is fully automated and no human feedback is required. Experiments conducted on a COREL image dataset demonstrate the effectiveness of the proposed approach.

Pairwise similarity of samples is an essential factor in graph propagation based semi-supervised learning methods. Usually it is estimated based on Euclidean distance. However, the structural assumption, which is a basic assumption in these methods, has not been taken into consideration in the normal pairwise similarity measure. In this paper, we propose a novel graph-based learning approach, named Structure-Sensitive Manifold Ranking (SSMR),based on a structure-sensitive similarity measure. Instead of using distance only, SSMR takes local distribution differences into account to more accurately measure pairwise similarity. Furthermore, we show that SSMR can also be deduced from a partial differential equation based anisotropic diffusion. Experiments conducted on the TRECVID dataset show that this approach significantly outperforms existing graph-based semi-supervised learning methods for video semantic concept detection.