The development of multi-view image synthesis is constrained by the scarcity of training data. One promising solution is to finetune well-trained video generative models to synthesize 360-degree videos of objects. While these methods benefit from the strong generative priors inherited from the pretrained knowledge, they are limited by the high computational costs incurred by the large number of viewpoints. Existing methods commonly adopt temporal attention mechanism to address this. However, these methods suffer from undesirable artifacts such as 3D inconsistency and over-smoothing in the generated results. In this paper, we introduce a novel approach to unlock the video priors for multi-view synthesis by reducing generation into a sparser yet more precise process. Specifically, we introduce two strategies to achieve this: i) Condensing the video diffusion model to synthesize highly consistent sparse multiview images. ii) Extracting dense geometrical priors from the pretrained video diffusion models to enhance the generation stability. The combination of these two strategies formulates a novel framework for multi-view synthesis, which is capable of synthesizing highly consistent sparse multiview images with strong generalization ability. Extensive experiments demonstrate that our approach achieves superior efficiency, generalization, and consistency, outperforming state-of-the-art multi-view synthesis methods.

LiDAR data generation has emerged as a promising solution to the high cost and limited scalability of real-world LiDAR sensing. Recent diffusion and rectified flow models have demonstrated strong capabilities in synthesizing realistic 3D point clouds; however, their iterative sampling procedures result in significant inference overhead. To address this, we focus on efficient few-step LiDAR generation for both unconditional and multi-modal conditional settings. Specifically, we propose an adaptive piecewise distillation strategy tailored for rectified flow-based LiDAR generation models, where the teacher model’s flow trajectory is adaptively segmented into consecutive intervals, and the student is trained only at the start of each interval to directly predict the velocity toward its endpoint. By sequentially sampling at the start timestep of each interval, our method enables fast few-step generation. Moreover, instead of uniform partitioning, we introduce an adaptive timestep selection strategy that chooses interval boundaries with minimal initial error, thereby reducing the complexity of distillation. Experimental results show that our method achieves comparable or superior performance to state-of-the-art methods in both unconditional and multi-modal conditional LiDAR generation, using only four sampling steps.

Understanding motion in dynamic environments is critical for autonomous driving, thereby motivating research on class-agnostic motion prediction. In this work, we investigate weakly and self-supervised class-agnostic motion prediction from LiDAR point clouds. Outdoor scenes typically consist of mobile foregrounds and static backgrounds, allowing motion understanding to be associated with scene parsing. Based on this observation, we propose a novel weakly supervised paradigm that replaces motion annotations with fully or partially annotated (1%, 0.1%) foreground/background masks for supervision. To this end, we develop a weakly supervised approach utilizing foreground/background cues to guide the self-supervised learning of motion prediction models. Since foreground motion generally occurs in non-ground regions, non-ground/ground masks can serve as an alternative to foreground/background masks, further reducing annotation effort. Leveraging non-ground/ground cues, we propose two additional approaches: a weakly supervised method requiring fewer (0.01%) foreground/background annotations, and a self-supervised method without annotations. Furthermore, we design a Robust Consistency-aware Chamfer Distance loss that incorporates multi-frame information and robust penalty functions to suppress outliers in self-supervised learning. Experiments show that our weakly and self-supervised models outperform existing self-supervised counterparts, and our weakly supervised models even rival some supervised ones. This demonstrates that our approaches effectively balance annotation effort and performance.

No-reference image quality assessment (NR-IQA) aims to predict perceptual image quality without access to pristine references, which remains challenging due to diverse and complex distortions. Recent pseudo-reference-based methods attempt to mitigate this challenge but often rely on highfidelity pseudo-reference reconstruction. In contrast, this work shows that improving NR-IQA performance does not depend on reconstruction quality, but on effective representation learning, feature alignment, and deviation modeling between distorted images and pseudo references. To this end, we propose a novel NR-IQA framework that leverages low-quality pseudo references generated by a masked autoencoder with a lightweight decoder. Rather than pursuing detailed reconstruction, the pseudo reference is used to facilitate representation-level deviation modeling in a shared latent space via a cross-attention-based mechanism. Extensive experiments on multiple benchmark datasets demonstrate that the proposed method consistently outperforms state-of-the-art NR-IQA approaches while maintaining modest computational complexity. Our source code will be available at: https://github.com/jianjin008/L-IQA.

Current methods for 3D semantic segmentation propose training models with limited annotations to address the difficulty of annotating large, irregular, and unordered 3D point cloud data. They usually focus on the 3D domain only, without leveraging the complementary nature of 2D and 3D data. Besides, some methods extend original labels or generate pseudo labels to guide the training, but they often fail to fully use these labels or address the noise within them. Meanwhile, the emergence of comprehensive and adaptable foundation models has offered effective solutions for segmenting 2D data. Leveraging this advancement, we present a novel approach that maximizes the utility of sparsely available 3D annotations by incorporating segmentation masks generated by 2D foundation models. We further propagate the 2D segmentation masks into the 3D space by establishing geometric correspondences between 3D scenes and 2D views. We extend the highly sparse annotations to encompass the areas delineated by 3D masks, thereby substantially augmenting the pool of available labels. Furthermore, we apply confidence- and uncertainty-based consistency regularization on augmentations of the 3D point cloud and select the reliable pseudo labels, which are further spread on the 3D masks to generate more labels. This innovative strategy bridges the gap between limited 3D annotations and the powerful capabilities of 2D foundation models, ultimately improving the performance of 3D weakly supervised segmentation.

Meta-learning has emerged as an effective solver for cross-domain few-shot learning (CD-FSL) tasks. Despite achieving obvious progress recently, the typical episodic learning paradigm often causes the feature embedding model collapsing into the simplicity bias pitfall, viz., the model tends to prioritize some shortcut patterns (e.g., color, style, background) that are only sufficient to distinguish categories in source domain, while fail to generalize across domains. To mitigate this problem, we present a novel meta-learning framework which emphasizes meta-exploiting inductive bias to alleviate simplicity bias for CD-FSL promotion, and mainly contributes in the following four aspects. 1) We establish a novel inductive bias for CD-FSL, termed complementary semantic consistency (CSC). The rationale behind lies in that forcing the semantic consistency between two complementary feature learning schemes is beneficial to distill cross-domain transferable features. 2) We establish a solid theoretical foundation, supported by rigorous mathematical proofs and key lemmas, which demonstrates that CSC establishes a tighter generalization bound and facilitates the learning of domain-invariant features. 3) Inspired by CSC, we propose a general meta-learning framework, which implements complementary feature embedding models using parallel networks with the same architecture but different input forms, and introduce proper knowledge distillation losses to encourage the semantic consistency between different branches during meta-training. This framework can be seamlessly integrated with any complementary feature learning schemes. 4) To clarify this point, we instantiate two effective meta-learners based on the proposed framework. The former establishes a two-branch network that simultaneously classifies both the query image and its random local crops. The latter decomposes the query image into high-frequency and low-frequency components, which are then integrated into a parallel feature embedding network for category prediction, analogous to the original query image. Subsequently, a KL divergence based knowledge distillation loss is separately leveraged to force the prediction consistency between the complementary branches (e.g., local-global, spatial-frequency) during meta-training. By doing these, both learners are able to distill cross-domain transferable features with better generalization performance. Empirical results on diverse benchmarks consistently affirm the proposed framework's advantages, while additional analysis provides compelling support for our key claims.

Multimodal sarcasm detection receives increasing attentions due to people's growing interest in posting multimodal information. The key factor of multimodal sarcasm detection is to leverage incongruity information across different modalities. Existing works are mainly based on the late fusion strategy by simply concatenating the intra- and inter-modal incongruity features, which are prone to learning surface patterns. In contrast, this work mainly focuses on modeling the deep fusion of intra- and inter-modal incongruity information. To this end, this work first discusses the incompatibility between the two kinds of incongruity features within existing multimodal frameworks. Under this motivation, we further propose an end-to-end cooperative framework dubbed Cooperative Multimodal Incongruity Learning (CoMIL). Specifically, our approach incorporates a primary module to model the deep fusion of intra- and inter modal incongruity information. To prevent the integrated inter modal visual information from disturbing the modeling of intra text incongruity, CoMIL introduces a cooperative mechanism incorporating a reference module which focuses on token-level correlations as a structural guidance to the primary module. Based on the proposed cooperative mechanism, the intra- and inter-modal incongruity information can be compactly and compatibly integrated into deep features of neural models. Extensive experiments are conducted to validate the effectiveness of our proposed CoMIL approach.

Given a pair of images depicting a person and a garment separately, image-based 3D virtual try-on methods aim to reconstruct a 3D human model that realistically portrays the person wearing the desired garment. In this paper, we present IPVTON, a novel image-based 3D virtual try-on framework. IPVTON employs score distillation sampling with image prompts to optimize a hybrid 3D human representation, integrating target garment features into diffusion priors through an image prompt adapter. To avoid interference with non-target areas, we leverage mask-guided image prompt embeddings to focus the image features on the try-on regions. Moreover, we impose geometric constraints on the 3D model with a pseudo silhouette generated by ControlNet, ensuring that the clothed 3D human model retains the shape of the source identity while accurately wearing the target garments. Extensive qualitative and quantitative experiments demonstrate that IPVTON outperforms previous methods in image-based 3D virtual try-on tasks, excelling in both geometry and texture.

Radar-Camera depth estimation aims to predict dense and accurate metric depth by fusing input images and Radar data. Model efficiency is crucial for this task in pursuit of real-time processing on autonomous vehicles and robotic platforms. However, due to the sparsity of Radar returns, the prevailing methods adopt multi-stage frameworks with intermediate quasi-dense depth, which are time-consuming and not robust. To address these challenges, we propose TacoDepth, an efficient and accurate Radar-Camera depth estimation model with one-stage fusion. Specifically, the graph-based Radar structure extractor and the pyramid-based Radar fusion module are designed to capture and integrate the graph structures of Radar point clouds, delivering superior model efficiency and robustness without relying on the intermediate depth results. Moreover, TacoDepth can be flexible for different inference modes, providing a better balance of speed and accuracy. Extensive experiments are conducted to demonstrate the efficacy of our method. Compared with the previous state-of-the-art approach, TacoDepth improves depth accuracy and processing speed by 12.8% and 91.8%. Our work provides a new perspective on efficient Radar-Camera depth estimation.

Creating CAD digital twins from the physical world is crucial for manufacturing, design, and simulation. However, current methods typically rely on costly 3D scanning with labor-intensive post-processing. To provide a user-friendly design process, we explore the problem of reverse engineering from unconstrained real-world CAD images that can be easily captured by users of all experiences. However, the scarcity of real-world CAD data poses challenges in directly training such models. To tackle these challenges, we propose CADCrafter, an image-to-parametric CAD model generation framework that trains solely on synthetic textureless CAD data while testing on real-world images. To bridge the significant representation disparity between images and parametric CAD models, we introduce a geometry encoder to accurately capture diverse geometric features. Moreover, the texture-invariant properties of the geometric features can also facilitate the generalization to real-world scenarios. Since compiling CAD parameter sequences into explicit CAD models is a non-differentiable process, the network training inherently lacks explicit geometric supervision. To impose geometric validity constraints, we employ direct preference optimization (DPO) to fine-tune our model with the automatic code checker feedback on CAD sequence quality. Furthermore, we collected a real-world dataset, comprised of multi-view images and corresponding CAD command sequence pairs, to evaluate our method. Experimental results demonstrate that our approach can robustly handle real unconstrained CAD images, and even generalize to unseen general objects.

With the explosive growth of 3D content creation, there is an increasing demand for automatically converting static 3D models into articulation-ready versions that support realistic animation. Traditional approaches rely heavily on manual annotation, which is both time-consuming and labor-intensive. Moreover, the lack of large-scale benchmarks has hindered the development of learning-based solutions. In this work, we present MagicArticulate, an effective framework that automatically transforms static 3D models into articulation-ready assets. Our key contributions are threefold. First, we introduce Articulation-XL, a large-scale benchmark containing over 33k 3D models with high-quality articulation annotations, carefully curated from Objaverse-XL. Second, we propose a novel skeleton generation method that formulates the task as a sequence modeling problem, leveraging an autoregressive transformer to naturally handle varying numbers of bones or joints within skeletons and their inherent dependencies across different 3D models. Third, we predict skinning weights using a functional diffusion process that incorporates volumetric geodesic distance priors between vertices and joints. Extensive experiments demonstrate that MagicArticulate significantly outperforms existing methods across diverse object categories, achieving high-quality articulation that enables realistic animation. Project page: https://chaoyuesong.github.io/MagicArticulate.

Language Bottleneck Models (LBMs) are proposed to achieve interpretable image recognition by classifying images based on textual concept bottlenecks. However, current LBMs simply list all concepts together as the bottleneck layer, leading to the spurious cue inference problem and cannot generalized to unseen classes. To address these limitations, we propose the Attribute-formed Language Bottleneck Model (ALBM). ALBM organizes concepts in the attribute-formed class-specific space, where concepts are descriptions of specific attributes for specific classes. In this way, ALBM can avoid the spurious cue inference problem by classifying solely based on the essential concepts of each class. In addition, the cross-class unified attribute set also ensures that the concept spaces of different classes have strong correlations, as a result, the learned concept classifier can be easily generalized to unseen classes. Moreover, to further improve interpretability, we propose Visual Attribute Prompt Learning (VAPL) to extract visual features on fine-grained attributes. Furthermore, to avoid labor-intensive concept annotation, we propose the Description, Summary, and Supplement (DSS) strategy to automatically generate high-quality concept sets with a complete and precise attribute. Extensive experiments on 9 widely used few-shot benchmarks demonstrate the interpretability, transferability, and performance of our approach. The code and collected concept sets are available at https://github.com/tiggers23/ALBM.

Meshes are the de facto 3D representation in the industry but are labor-intensive to produce. Recently, a line of research has focused on autoregressively generating meshes. This approach processes meshes into a sequence composed of vertices and then generates them vertex by vertex, similar to how a language model generates text. These methods have achieved some success but still struggle to generate complex meshes. One primary reason for this limitation is their inefficient tokenization methods. To address this issue, we introduce MeshAnything V2, an advanced mesh generation model designed to create Artist-Created Meshes that align precisely with specified shapes. A key innovation behind MeshAnything V2 is our novel Adjacent Mesh Tokenization (AMT) method. Unlike traditional approaches that represent each face using three vertices, AMT optimizes this by employing a single vertex wherever feasible, effectively reducing the token sequence length by about half on average. This not only streamlines the tokenization process but also results in more compact and well-structured sequences, enhancing the efficiency of mesh generation. With these improvements, MeshAnything V2 effectively doubles the face limit compared to previous models, delivering superior performance without increasing computational costs. We will make our code and models publicly available. Project Page: https://buaacyw.github.io/meshanything-v2/

Surface reconstruction and novel view rendering from sparse-view images are challenging. Signed Distance Function (SDF)-based methods struggle with fine details, while 3D Gaussian Splatting (3DGS)-based approaches lack global geometry coherence. We propose a novel hybrid method that combines the strengths of both approaches: SDF captures coarse geometry to enhance 3DGS-based rendering, while newly rendered images from 3DGS refine the details of SDF for accurate surface reconstruction. As a result, our method surpasses state-of-the-art approaches in surface reconstruction and novel view synthesis on the DTU and MobileBrick datasets. Code will be released at https://github.com/aim-uofa/SurfaceSplat.

We propose MotionAgent, enabling fine-grained motion control for text-guided image-to-video generation. The key technique is the motion field agent that converts motion information in text prompts into explicit motion fields, providing flexible and precise motion guidance. Specifically, the agent extracts the object movement and camera motion described in the text and converts them into object trajectories and camera extrinsics, respectively. An analytical optical flow composition module integrates these motion representations in 3D space and projects them into a unified optical flow. An optical flow adapter takes the flow to control the base image-to-video diffusion model for generating fine-grained controlled videos. The significant improvement in the Video-Text Camera Motion metrics on VBench indicates that our method achieves precise control over camera motion. We construct a subset of VBench to evaluate the alignment of motion information in the text and the generated video, outperforming other advanced models on motion generation accuracy.

Panoptic segmentation, which unifies semantic and instance segmentation into a single task, has witnessed considerable success on predefined tasks. However, traditional methods tend to struggle with catastrophic forgetting and poor generalization when learning from a continuous stream of new tasks. While continual learning aims to mitigate these challenges, our study reveals that existing continual panoptic segmentation (CPS) methods often suffer from efficiency or scalability issues. To address these limitations, we propose an efficient adaptation framework that incorporates attentive self-distillation and dual-decoder prediction fusion to efficiently preserve prior knowledge while facilitating model generalization. Specifically, we freeze the majority of model weights, enabling a shared forward pass between the teacher and student models during distillation. Attentive self-distillation then adaptively distills useful knowledge from the old classes without being distracted from non-object regions, which effectively enhances knowledge retention. Additionally, query-level fusion (QLF) is devised to seamlessly integrate the output of the dual decoders without incurring scale inconsistency. Our method achieves state-of-the-art performance on ADE20K and COCO benchmarks. Code is available at https://github.com/Ze-Yang/ADAPT.

Existing methods for 3D clothed human digitalization perform well when the input image is captured in ideal conditions that assume the lack of any occlusion. However, in reality, images may often have occlusion problems such as incomplete observation of the human subject's full body, self-occlusion by the human subject, and non-frontal body pose. When given such input images, these existing methods fail to perform adequately. Thus, we propose Robust-PIFu, a pixel-aligned implicit model that capitalized on large-scale, pretrained latent diffusion models to address the challenge of digitalizing human subjects from non-ideal images that suffer from occlusions. Robust-PIfu offers four new contributions. Firstly, we propose a 'disentangling' latent diffusion model. This diffusion model, pretrained on billions of images, takes in any input image and removes external occlusions, such as inter-person occlusions, from that image. Secondly, Robust-PIFu addresses internal occlusions like self-occlusion by introducing a `penetrating' latent diffusion model. This diffusion model outputs multi-layered normal maps that by-pass occlusions caused by the human subject's own limbs or other body parts (i.e. self-occlusion). Thirdly, in order to incorporate such multi-layered normal maps into a pixel-aligned implicit model, we introduce our Layered-Normals Pixel-aligned Implicit Model, which improves the structural accuracy of predicted clothed human meshes. Lastly, Robust-PIFu proposes an optional super-resolution mechanism for the multi-layered normal maps. This addresses scenarios where the input image is of low or inadequate resolution. Though not strictly related to occlusion, this is still an important subproblem. Our experiments show that Robust-PIFu outperforms current SOTA methods both qualitatively and quantitatively. Our code will be released to the public.

Recently, 3D assets created via reconstruction and generation have matched the quality of manually crafted assets, highlighting their potential for replacement. However, this potential is largely unrealized because these assets always need to be converted to meshes for 3D industry applications, and the meshes produced by current mesh extraction methods are significantly inferior to Artist-Created Meshes (AMs), i.e., meshes created by human artists. Specifically, current mesh extraction methods rely on dense faces and ignore geometric features, leading to inefficiencies, complicated post-processing, and lower representation quality. To address these issues, we introduce MeshAnything, a model that treats mesh extraction as a generation problem, producing AMs aligned with specified shapes. By converting 3D assets in any 3D representation into AMs, MeshAnything can be integrated with various 3D asset production methods, thereby enhancing their application across the 3D industry. The architecture of MeshAnything comprises a VQ-VAE and a shape-conditioned decoder-only transformer. We first learn a mesh vocabulary using the VQ-VAE, then train the shape-conditioned decoder-only transformer on this vocabulary for shape-conditioned autoregressive mesh generation. Our extensive experiments show that our method generates AMs with hundreds of times fewer faces, significantly improving storage, rendering, and simulation efficiencies, while achieving precision comparable to previous methods.

Large-scale diffusion models have achieved remarkable performance in generative tasks. Beyond their initial training applications, these models have proven their ability to function as versatile plug-and-play priors. For instance, 2D diffusion models can serve as loss functions to optimize 3D implicit models. Rectified Flow, a novel class of generative models, has demonstrated superior performance across various domains. Compared to diffusion-based methods, rectified flow approaches surpass them in terms of generation quality and efficiency. In this work, we present theoretical and experimental evidence demonstrating that rectified flow based methods offer similar functionalities to diffusion models — they can also serve as effective priors. Besides the generative capabilities of diffusion priors, motivated by the unique time-symmetry properties of rectified flow models, a variant of our method can additionally perform image inversion. Experimentally, our rectified flow based priors outperform their diffusion counterparts — the SDS and VSD losses — in text-to-3D generation. Our method also displays competitive performance in image inversion and editing. Code is available at: https://github.com/yangxiaofeng/rectified_flow_prior.

Human mesh recovery (HMR) from a single image is inherently ill-posed due to depth ambiguity and occlusions. Probabilistic methods have tried to solve this by generating numerous plausible 3D human mesh predictions, but they often exhibit misalignment with 2D image observations and weak robustness to in-the-wild images. To address these issues, we propose ADHMR, a framework that Aligns a Diffusion-based HMR model in a preference optimization manner. First, we train a human mesh prediction assessment model, HMR-Scorer, capable of evaluating predictions even for in-the-wild images without 3D annotations. We then use HMR-Scorer to create a preference dataset, where each input image has a pair of winner and loser mesh predictions. This dataset is used to finetune the base model using direct preference optimization. Moreover, HMR-Scorer also helps improve existing HMR models by data cleaning, even with fewer training samples. Extensive experiments show that ADHMR outperforms current state-of-the-art methods. Code is available at: https://github.com/shenwenhao01/ADHMR.

Few-Shot Segmentation (FSS) aims to learn class-agnostic segmentation on few classes to segment arbitrary classes, but at the risk of overfitting. To address this, some methods use the well-learned knowledge of foundation models (e.g., SAM) to simplify the learning process. Recently, SAM 2 has extended SAM by supporting video segmentation, whose class-agnostic matching ability is useful to FSS. A simple idea is to encode support foreground (FG) features as memory, with which query FG features are matched and fused. Unfortunately, the FG objects in different frames of SAM 2’s video data are always the same identity, while those in FSS are different identities, i.e., the matching step is incompatible. Therefore, we design Pseudo Prompt Generator to encode pseudo query memory, matching with query features in a compatible way. However, the memories can never be as accurate as the real ones, i.e., they are likely to contain incomplete query FG, and some unexpected query background (BG) features, leading to wrong segmentation. Hence, we further design Iterative Memory Refinement to fuse more query FG features into the memory, and devise a Support-Calibrated Memory Attention to suppress the unexpected query BG features in memory. Extensive experiments have been conducted on PASCAL-5$^i$ and COCO-20$^i$ to validate the effectiveness of our design, e.g., the 1-shot mIoU can be 4.2% better than the best baseline.

Robot navigation in dense human crowds poses a significant challenge due to the complexity of human behavior in dynamic and obstacle-rich environments. In this work, we propose a dynamic weight adjustment scheme using a neural network to predict the optimal weights of objectives in an optimization-based motion planner. We adopt a spatial-temporal trajectory planner and incorporate diverse objectives to achieve a balance among safety, efficiency, and goal achievement in complex and dynamic environments. We design the network structure, observation encoding, and reward function to effectively train the policy network using reinforcement learning, allowing the robot to adapt its behavior in real time based on environmental and pedestrian information. Simulation results show improved safety compared to the fixed-weight planner and the state-of-the-art learning-based methods, and verify the ability of the learned policy to adaptively adjust the weights based on the observed situations. The feasibility of the approach is demonstrated in a navigation task using an autonomous delivery robot across a crowded corridor over a 300 m distance. Video: https://youtu.be/nSCbNaaF_VM

In this paper, we investigate an open research task of generating 3D cartoon face shapes from single 2D GAN generated human faces and without 3D supervision, where we can also manipulate the facial expressions of the 3D shapes. To this end, we discover the semantic meanings of StyleGAN latent space, such that we are able to produce face images of various expressions, poses, and lighting conditions by controlling the latent codes. Specifically, we first finetune the pretrained StyleGAN face model on the cartoon datasets. By feeding the same latent codes to face and cartoon generation models, we aim to realize the translation from 2D human face images to cartoon styled avatars. We then discover semantic directions of the GAN latent space, in an attempt to change the facial expressions while preserving the original identity. As we do not have any 3D annotations for cartoon faces, we manipulate the latent codes to generate images with different poses and lighting conditions, such that we can reconstruct the 3D cartoon face shapes. We validate the efficacy of our method on three cartoon datasets qualitatively and quantitatively.

Controllable 3D indoor scene synthesis stands at the forefront of technological progress, offering various applications like gaming, film, and augmented/virtual reality. The capability to stylize and decouple objects within these scenarios is a crucial factor, providing an advanced level of control throughout the editing process. This control extends not just to manipulating geometric attributes like translation and scaling but also includes managing appearances, such as stylization. Current methods for scene stylization are limited to applying styles to the entire scene, without the ability to separate and customize individual objects. Addressing the intricacies of this challenge, we introduce a unique pipeline designed for synthesis 3D indoor scenes. Our approach involves strategically placing objects within the scene, utilizing information from professionally designed bounding boxes. Significantly, our pipeline prioritizes maintaining style consistency across multiple objects within the scene, ensuring a cohesive and visually appealing result aligned with the desired aesthetic. The core strength of our pipeline lies in its ability to generate 3D scenes that are not only visually impressive but also exhibit features like photo-realistic, multi-view consistency, and diversity, demonstrating the versatility and adaptability of our model.

Controllable 3D indoor scene synthesis stands at the forefront of technological progress, offering various applications like gaming, film, and augmented/virtual reality. The capability to stylize and decouple objects within these scenarios is a crucial factor, providing an advanced level of control throughout the editing process. This control extends not just to manipulating geometric attributes like translation and scaling but also includes managing appearances, such as stylization. Current methods for scene stylization are limited to applying styles to the entire scene, without the ability to separate and customize individual objects. Addressing the intricacies of this challenge, we introduce a unique pipeline designed for synthesis 3D indoor scenes. Our approach involves strategically placing objects within the scene, utilizing information from professionally designed bounding boxes. Significantly, our pipeline prioritizes maintaining style consistency across multiple objects within the scene, ensuring a cohesive and visually appealing result aligned with the desired aesthetic. The core strength of our pipeline lies in its ability to generate 3D scenes that are not only visually impressive but also exhibit features like photo-realistic, multi-view consistency, and diversity, demonstrating the versatility and adaptability of our model.

The widespread deployment of large models in resource-constrained environments has underscored the need for efficient transmission of intermediate feature representations. In this context, feature coding, which compresses features into compact bitstreams, becomes a critical component for scenarios involving feature transmission, storage, and reuse. However, this compression process inevitably introduces semantic degradation that is difficult to quantify with traditional metrics. To address this, we formalize the research problem of Compressed Feature Quality Assessment (CFQA), aiming to evaluate the semantic fidelity of compressed features. To advance CFQA research, we propose the first benchmark dataset, comprising 300 original features and 12000 compressed features derived from three vision tasks and four feature codecs. Task-specific performance degradation is provided as true semantic distortion for evaluating CFQA metrics. We systematically assess three widely used metrics -- MSE, cosine similarity, and Centered Kernel Alignment (CKA) -- in terms of their ability to capture semantic degradation. Our findings demonstrate the representativeness of the proposed dataset while underscoring the need for more sophisticated metrics capable of measuring semantic distortion in compressed features. This work advances the field by establishing a foundational benchmark and providing a critical resource for the community to explore CFQA. To foster further research, we release the dataset and all associated source code at https://github.com/chansongoal/Compressed-Feature-Quality-Assessment.

While existing research on Multilingual CLIP (MCLIP) has prioritized model architecture design, our work uncovers a critical challenge in practical adaptation: fine-tuning MCLIP through a single source language risks diminishing its multilingual capabilities in downstream tasks due to cross-linguistic disparities. To bridge this gap, we systematically investigate the role of ‌token similarity‌ in cross-lingual transferability for image-text retrieval, establishing it as a key factor governing fine-tuning efficacy. Building on this insight, we propose two novel strategies to enhance efficiency while preserving multilinguality: 1) ‌TaPCL dynamically optimizes training by prioritizing linguistically distant language pairs during corpus sampling, reducing redundant computation, and 2) CiPCL enriches the source corpus with multilingual key terms, enabling targeted knowledge transfer without reliance on exhaustive parallel data. By strategically balancing token similarity and domain-critical information, our methods significantly lower computational costs and mitigate over-dependence on parallel corpora. Experimental evaluations across diverse datasets validate the effectiveness and scalability of our framework, demonstrating robust multilingual retention across languages. This work provides a principled pathway for adapting MCLIP to real-world scenarios, where computational efficiency and cross-lingual robustness are paramount. Our codes are available at https://github.com/tiggers23/TaPCL-CiPCL.

Modern interactive applications increasingly demand dynamic 3D content, yet the transformation of static 3D models into animated assets constitutes a significant bottleneck in content creation pipelines. While recent advances in generative AI have revolutionized static 3D model creation, rigging and animation continue to depend heavily on expert intervention. We present \textbf{Puppeteer}, a comprehensive framework that addresses both automatic rigging and animation for diverse 3D objects. Our system first predicts plausible skeletal structures via an auto-regressive transformer that introduces a joint-based tokenization strategy for compact representation and a hierarchical ordering methodology with stochastic perturbation that enhances bidirectional learning capabilities. It then infers skinning weights via an attention-based architecture incorporating topology-aware joint attention that explicitly encodes inter-joint relationships based on skeletal graph distances. Finally, we complement these rigging advances with a differentiable optimization-based animation pipeline that generates stable, high-fidelity animations while being computationally more efficient than existing approaches. Extensive evaluations across multiple benchmarks demonstrate that our method significantly outperforms state-of-the-art techniques in both skeletal prediction accuracy and skinning quality. The system robustly processes diverse 3D content, ranging from professionally designed game assets to AI-generated shapes, producing temporally coherent animations that eliminate the jittering issues common in existing methods.

We present the first framework capable of synthesizing the all-in-focus neural radiance field (NeRF) from inputs without manual refocusing. Without refocusing, the camera will automatically focus on the fixed object for all views, and current NeRF methods typically using one camera fail due to the consistent defocus blur and a lack of sharp reference. To restore the all-in-focus NeRF, we introduce the dual-camera from smartphones, where the ultra-wide camera has a wider depth-of-field (DoF) and the main camera possesses a higher resolution. The dual camera pair saves the high-fidelity details from the main camera and uses the ultra-wide camera's deep DoF as reference for all-in-focus restoration. To this end, we first implement spatial warping and color matching to align the dual camera, followed by a defocus-aware fusion module with learnable defocus parameters to predict a defocus map and fuse the aligned camera pair. We also build a multi-view dataset that includes image pairs of the main and ultra-wide cameras in a smartphone. Extensive experiments on this dataset verify that our solution, termed DC-NeRF, can produce high-quality all-in-focus novel views and compares favorably against strong baselines quantitatively and qualitatively. We further show DoF applications of DC-NeRF with adjustable blur intensity and focal plane, including refocusing and split diopter.

Despite significant advancements in simulating the bokeh effect of Digital Single Lens Reflex Camera (DSLR) from an all-in-focus image, challenges remain in processing highlight points, preserving boundary details for in-focus objects and processing high-resolution images efficiently. To tackle these issues, we first develop a ray-tracing-based bokeh simulator. An innovative pipeline with weight redistribution is introduced to handle highlight rendering. By considering the front length of lens barrel, we can simulate realistic cat-eye effect. This bokeh simulator serves as the foundation for creating our training dataset. Building on this dataset, we introduce a hybrid framework BokehMe++, combining a classical renderer and a neural renderer. The classical renderer is implemented by a hierarchical scattering-based method, which suffers from boundary inaccuracies. These erroneous areas will be identified by an error map generator and be corrected by a two-stage neural renderer. Adaptive resizing and iterative upsampling are introduced in the neural renderer to process arbitrary blur size efficiently. Extensive experiments demonstrate that BokehMe++ outperforms existing methods and provides highly customizable rendering features, such as adjustable blur amount, focal plane, highlight mode and cat-eye effect. Furthermore, BokehMe++ can maintain the sharpness of hair details in portraits through an auxiliary alpha map input.

Recently, some weakly supervised 3D point cloud segmentation methods have been proposed to develop effective models with minimum annotation efforts. Our previous work, W4DTS, proposes a challenging task that utilizes only 0.001% points in outdoor point cloud datasets to achieve an effective segmentation model. However, under an extremely limited annotation budget, the quality of pseudo labels generated by W4DTS is unsatisfactory, which limits the segmentation performance in such scenarios. To solve this issue, we propose a progressive 4D grouping approach to group the annotated and unannotated points across space and time, which can generate high-quality pseudo labels with very sparse annotated points. Moreover, to further improve our progressive 4D grouping approach, we design a cross-frame contrastive learning and a local consistency learning to improve the quality of our 4D grouping. Experimental results reveal that with only 0.001% annotations, our solution significantly outperforms the previous best approach on SemanticKITTI. We also evaluate our framework on the SemanticPOSS dataset and ScribbleKITTI dataset, and achieve performances close to our fully supervised backbone models.

Video depth estimation aims to infer temporally consistent depth. One approach is to finetune a single-image model on each video with geometry constraints, which proves inefficient and lacks robustness. An alternative is learning to enforce consistency from data, which requires well-designed models and sufficient video depth data. To address both challenges, we introduce NVDS that stabilizes inconsistent depth estimated by various single-image models in a plug-and-play manner. We also elaborate a large-scale Video Depth in the Wild (VDW) dataset, which contains 14,203 videos with over two million frames, making it the largest natural-scene video depth dataset. Additionally, a bidirectional inference strategy is designed to improve consistency by adaptively fusing forward and backward predictions. We instantiate a model family ranging from small to large scales for different applications. The method is evaluated on VDW dataset and three public benchmarks. To further prove the versatility, we extend NVDS to video semantic segmentation and several downstream applications like bokeh rendering, novel view synthesis, and 3D reconstruction. Experimental results show that our method achieves significant improvements in consistency, accuracy, and efficiency. Our work serves as a solid baseline and data foundation for learning-based video depth estimation.

Recent neural networks based surface reconstruction can be roughly divided into two categories, one warping templates explicitly and the other representing 3D surfaces implicitly. To enjoy the advantages of both, we propose a novel 3D representation, Neural Vector Fields (NVF), which adopts the explicit learning process to manipulate meshes and implicit unsigned distance function (UDF) representation to break the barriers in resolution and topology. This is achieved by directly predicting the displacements from surface queries and modeling shapes as Vector Fields, rather than relying on network differentiation to obtain direction fields as most existing UDF-based methods do. In this way, our approach is capable of encoding both the distance and the direction fields so that the calculation of direction fields is differentiation-free, circumventing the non-trivial surface extraction step. Furthermore, building upon NVFs, we propose to incorporate two types of shape codebooks, i.e., NVFs (Lite or Ultra), to promote cross-category reconstruction through encoding cross-object priors. Moreover, we propose a new regularization based on analyzing the zero-curl property of NVFs, and implement this through the fully differentiable framework of our NVF (ultra). We evaluate both NVFs on four surface reconstruction scenarios, including watertight vs non-watertight shapes, category-agnostic reconstruction vs category-unseen reconstruction, category-specific, and cross-domain reconstruction.

Most existing Domain Generalizable Person Re-identification (DG-ReID) methods focus on addressing style disparities between domains but often overlook the impact of unpredictable camera view changes, which we have identified as a significant factor responsible for poor generalization performance. To address this issue, we propose a novel approach from a 3D perspective, utilizing a customized 2D-to-3D reconstruction model to convert images captured from arbitrary camera views into canonical view images. However, merely applying a 3D reconstruction model in isolation may not result in improved DG-ReID performance, as reconstruction quality can be influenced by multiple factors, such as insufficient image resolution, extreme viewpoint, and environmental variations. These factors may lead to error accumulation and the loss of critical discriminative clues in the reconstructed results. To address this difficulty, we propose fusing the canonical view image with the original image using a transformer-based module. The transformer’s cross-attention mechanism is ideal for aligning and fusing the key semantic clues of the original image with the canonical view image, compensating for reconstruction errors. We demonstrate the effectiveness of our method through extensive experiments in various evaluation settings, achieving superior DG-ReID performance compared to existing approaches. Our approach addresses the impact of unpredictable camera view changes and provides a new perspective for designing DG-ReID methods.

Training a generative model with limited data (e.g., 10) is a very challenging task. Many works propose to fine-tune a pretrained GAN model. However, this can easily result in overfitting. In other words, they manage to adapt the style but fail to preserve the content, where style denotes the specific properties that define a domain while content denotes the domain-irrelevant information that represents diversity. Recent works try to maintain a predefined correspondence to preserve the content, however, the diversity is still not enough and it may affect style adaptation. In this work, we propose a paired image reconstruction approach for content preservation. We propose to introduce an image translation module to GAN transferring, where the module teaches the generator to separate style and content, and the generator provides training data to the translation module in return. Qualitative and quantitative experiments show that our method consistently surpasses the state-of-the-art methods in a few-shot setting.

Few-shot single-view 3D reconstruction learns to reconstruct the novel category objects based on a query image and a few support shapes. However, since the query image and the support shapes are of different modalities, there is an inherent feature misalignment problem damaging the reconstruction. Previous works in the literature do not consider this problem. To this end, we propose the cross-modal feature alignment network (CMFAN) with two novel techniques. One is a strategy for model pretraining, namely, cross-modal contrastive learning (CMCL), here the 2D images and 3D shapes of the same objects compose the positives, and those from different objects form the negatives. With CMCL, the model learns to embed the 2D and 3D modalities of the same object into a tight area in the feature space and push away those from different objects, thus effectively aligning the global cross-modal features. The other is cross-modal feature fusion (CMFF), which further aligns and fuses the local features. Specifically, it first re-represents the local features with the cross-attention operation, making the local features share more information. Then, CMFF generates a descriptor for the support features and attaches it to each local feature vector of the query image with dense concatenation. Moreover, CMFF can be applied to multilevel local features and brings further advantages. We conduct extensive experiments to evaluate the effectiveness of our designs, and CMFAN sets new state-of-the-art performance in all of the 1-/10-/25-shot tasks of ShapeNet and ModelNet datasets.

Reverse engineering CAD models from raw geometry is a classic but challenging research problem. In particular, reconstructing the CAD modeling sequence from point clouds provides great interpretability and convenience for editing. Analyzing previous work, we observed that a CAD modeling sequence represented by tokens and processed by a generative model does not have an immediate geometric interpretation. To improve upon this problem, we introduce geometric guidance into the reconstruction network. Our proposed model, PS-CAD, reconstructs the CAD modeling sequence one step at a time as illustrated in Figure 1 . At each step, we provide three forms of geometric guidance. First, we provide the geometry of surfaces where the current reconstruction differs from the complete model as a point cloud. This helps the framework to focus on regions that still need work. Second, we use geometric analysis to extract a set of planar prompts, that correspond to candidate surfaces where a CAD extrusion step could be started. Third, we present a step-wise sampling to generate multiple complete candidate CAD modeling steps instead of single-tokens without direct geometric interpretation. Our framework has three major components. Geometric guidance computation extracts the first two types of geometric guidance. Single-step reconstruction computes a single candidate CAD modeling step for each provided prompt. Single-step selection selects among the candidate CAD modeling steps. The process continues until the reconstruction is completed. Our quantitative results show a significant improvement across all metrics. For example, on the dataset DeepCAD, PS-CAD improves upon the best published SOTA method by reducing the geometry errors (CD and HD) by 10%, and the structural error (ECD metric) by about 13%.

Novel view synthesis for dynamic 3D scenes poses a significant challenge. Many notable efforts use NeRF-based approaches to address this task and yield impressive results. However, these methods rely heavily on sufficient motion parallax in the input images or videos. When the camera motion range becomes limited or even stationary (i.e., small camera motion), existing methods encounter two primary challenges: incorrect representation of scene geometry and inaccurate estimation of camera parameters. These challenges make prior methods struggle to produce satisfactory results or even become ineffective. To address the first challenge, we propose a novel Distribution-based Depth Regularization (DDR) that ensures the rendering weight distribution to align with the true distribution. Specifically, unlike previous methods that use depth loss to calculate the error of the expectation, we calculate the expectation of the error by using Gumbel-softmax to differentiably sample points from discrete rendering weight distribution. Additionally, we introduce constraints that enforce the volume density of spatial points before the object boundary along the ray to be near zero, ensuring that our model learns the correct geometry of the scene. To demystify the DDR, we further propose a visualization tool that enables observing the scene geometry representation at the rendering weight level. For the second challenge, we incorporate camera parameter learning during training to enhance the robustness of our model to camera parameters. We conduct extensive experiments to demonstrate the effectiveness of our approach in representing scenes with small camera motion input, and our results compare favorably to state-of-the-art methods.

Pixel-aligned implicit models, such as PIFu, PIFuHD, and ICON, are used for single-view clothed human reconstruction. These models need to be trained using a sampling training scheme. Existing sampling training schemes either fail to capture thin surfaces (e.g. ears, fingers) or cause noisy artefacts in reconstructed meshes. To address these problems, we introduce Fine Structured-Aware Sampling (FSS), a new sampling training scheme to train pixel-aligned implicit models for single-view human reconstruction. FSS resolves the aforementioned problems by proactively adapting to the thickness and complexity of surfaces. In addition, unlike existing sampling training schemes, FSS shows how normals of sample points can be capitalized in the training process to improve results. Lastly, to further improve the training process, FSS proposes a mesh thickness loss signal for pixel-aligned implicit models. It becomes computationally feasible to introduce this loss once a slight reworking of the pixel-aligned implicit function framework is carried out. Our results show that our methods significantly outperform SOTA methods qualitatively and quantitatively. Our code is publicly available at https://github.com/kcyt/FSS.

Recent strides in Text-to-3D techniques have been propelled by distilling knowledge from powerful large text-to-image diffusion models (LDMs). Nonetheless, existing Text-to-3D approaches often grapple with challenges such as over-saturation, inadequate detailing, and unrealistic outputs. This study presents a novel strategy that leverages explicitly synthesized multi-view images to address these issues. Our approach involves the utilization of image-to-image pipelines, empowered by LDMs, to generate posed high-quality images based on the renderings of coarse 3D models. Although the generated images mostly alleviate the aforementioned issues, challenges such as view inconsistency and significant content variance persist due to the inherent generative nature of large diffusion models, posing extensive difficulties in leveraging these images effectively. To overcome this hurdle, we advocate integrating a discriminator alongside a novel Diffusion-GAN dual training strategy to guide the training of 3D models. For the incorporated discriminator, the synthesized multi-view images are considered real data, while the renderings of the optimized 3D models function as fake data. We conduct a comprehensive set of experiments that demonstrate the effectiveness of our method over baseline approaches.

Class-agnostic motion prediction methods aim to comprehend motion within open-world scenarios, holding significance for autonomous driving systems. However, training a high-performance model in a fully-supervised manner always requires substantial amounts of manually annotated data, which can be both expensive and time-consuming to obtain. To address this challenge, our study explores the potential of semi-supervised learning (SSL) for class-agnostic motion prediction. Our SSL framework adopts a consistency-based self-training paradigm, enabling the model to learn from unlabeled data by generating pseudo labels through test-time inference. To improve the quality of pseudo labels, we propose a novel motion selection and re-generation module. This module effectively selects reliable pseudo labels and re-generates unreliable ones. Furthermore, we propose two data augmentation strategies: temporal sampling and BEVMix. These strategies facilitate consistency regularization in SSL. Experiments conducted on nuScenes demonstrate that our SSL method can surpass the self-supervised approach by a large margin by utilizing only a tiny fraction of labeled data. Furthermore, our method exhibits comparable performance to weakly and some fully supervised methods. These results highlight the ability of our method to strike a favorable balance between annotation costs and performance. Code will be available at https://github.com/kwwcv/SSMP.

In recent years, following the success of text guided image generation, text guided 3D generation has gained increasing attention among researchers. Dreamfusion is a notable approach that enhances generation quality by utilizing 2D text guided diffusion models and introducing SDS loss, a technique for distilling 2D diffusion model information to train 3D models. However, the SDS loss has two major limitations that hinder its effectiveness. Firstly, when given a text prompt, the SDS loss struggles to produce diverse content. Secondly, during training, SDS loss may cause the generated content to overfit and collapse, limiting the model's ability to learn intricate texture details. To overcome these challenges, we propose a novel approach called Noise Recalibration algorithm. By incorporating this technique, we can generate 3D content with significantly greater diversity and stunning details. Our approach offers a promising solution to the limitations of SDS loss.

Recently, the authors of Zero-1-to-3 demonstrated that a latent diffusion model, pretrained with Internet-scale data, can not only address the single-view 3D object reconstruction task but can even attain SOTA results in it. However, when applied to the task of single-view 3D clothed human reconstruction, Zero-1-to-3 (and related models) are unable to compete with the corresponding SOTA methods in this field despite being trained on clothed human data. In this work, we aim to tailor Zero-1-to-3's approach to the single-view 3D clothed human reconstruction task in a much more principled and structured manner. To this end, we propose R-Cyclic Diffuser, a framework that adapts Zero-1-to-3's novel approach to clothed human data by fusing it with a pixel-aligned implicit model. R-Cyclic Diffuser offers a total of three new contributions. The first and primary contribution is R-Cyclic Diffuser's cyclical conditioning mechanism for novel view synthesis. This mechanism directly addresses the view inconsistency problem faced by Zero-1-to-3 and related models. Secondly, we further enhance this mechanism with two key features - Lateral Inversion Constraint and Cyclic Noise Selection. Both features are designed to regularize and restrict the randomness of outputs generated by a latent diffusion model. Thirdly, we show how SMPL-X body priors can be incorporated in a latent diffusion model such that novel views of clothed human bodies can be generated much more accurately. Our experiments show that R-Cyclic Diffuser is able to outperform current SOTA methods in singleview 3D clothed human reconstruction both qualitatively and quantitatively. Our code is made publicly available at https://github.com/kcyt/r-cyclic-diffuser.

3D editing plays a crucial role in many areas such as gaming and virtual reality. Traditional 3D editing methods, which rely on representations like meshes and point clouds, often fall short in realistically depicting complex scenes. On the other hand, methods based on implicit 3D representations, like Neural Radiance Field (NeRF), render complex scenes effectively but suffer from slow processing speeds and limited control over specific scene areas. In response to these challenges, our paper presents GaussianEditor, the first 3D editing algorithm based on Gaussian Splatting (GS), a novel 3D representation. GaussianEditor enhances precision and control in editing through our proposed Gaussian semantic tracing, which traces the editing target throughout the training process. Additionally, we propose Hierarchical Gaussian splatting (HGS) to achieve stabilized and fine results under stochastic generative guidance from 2D diffusion models. We also develop editing strategies for efficient object removal and integration, a challenging task for existing methods. Our comprehensive experiments demonstrate GaussianEditor's superior control, effective, and efficient performance, marking a significant advancement in 3D editing.

Recent works on text-to-3d generation show that using only 2D diffusion supervision for 3D generation tends to produce results with inconsistent appearances (e.g., faces on the back view) and inaccurate shapes (e.g., animals with extra legs). Existing methods mainly address this issue by retraining diffusion models with images rendered from 3D data to ensure multi-view consistency while struggling to balance 2D generation quality with 3D consistency. In this paper, we present a new framework Sculpt3D that equips the current pipeline with explicit injection of 3D priors from retrieved reference objects without re-training the 2D diffusion model. Specifically, we demonstrate that high-quality and diverse 3D geometry can be guaranteed by keypoints supervision through a sparse ray sampling approach. Moreover, to ensure accurate appearances of different views, we further modulate the output of the 2D diffusion model to the correct patterns of the template views without altering the generated object's style. These two decoupled designs effectively harness 3D information from reference objects to generate 3D objects while preserving the generation quality of the 2D diffusion model. Extensive experiments show our method can largely improve the multi-view consistency while retaining fidelity and diversity. Our project page is available at: https://stellarcheng.github.io/Sculpt3D/.

Current 3D stylization methods often assume static scenes, which violates the dynamic nature of our real world. To address this limitation, we present S-DyRF, a reference-based spatio-temporal stylization method for dynamic neu-ral radiance fields. However, stylizing dynamic 3D scenes is inherently challenging due to the limited availability of stylized reference images along the temporal axis. Our key insight lies in introducing additional temporal cues besides the provided reference. To this end, we generate temporal pseudo-references from the given stylized reference. These pseudo-references facilitate the propagation of style infor-mation from the reference to the entire dynamic 3D scene. For coarse style transfer, we enforce novel views and times to mimic the style details present in pseudo-references at the feature level. To preserve high-frequency details, we create a collection of stylized temporal pseudo-rays from temporal pseudo-references. These pseudo-rays serve as detailed and explicit stylization guidance for achieving fine style trans-fer. Experiments on both synthetic and real-world datasets demonstrate that our method yields plausible stylized re-sults of space-time view synthesis on dynamic 3D scenes.

In this paper, we address the challenge of reconstructing general articulated 3D objects from a single video. Existing works employing dynamic neural radiance fields have advanced the modeling of articulated objects like humans and animals from videos, but face challenges with piece-wise rigid general articulated objects due to limitations in their deformation models. To tackle this, we propose Quasi-Rigid Blend Skinning, a novel deformation model that enhances the rigidity of each part while maintaining flexible deformation of the joints. Our primary insight combines three distinct approaches: 1) an enhanced bone rigging system for improved component modeling, 2) the use of quasi-sparse skinning weights to boost part rigidity and reconstruction fidelity, and 3) the application of geodesic point assignment for precise motion and seamless deformation. Our method outperforms previous works in producing higher-fidelity 3D reconstructions of general articulated objects, as demonstrated on both real and synthetic datasets. Project page: https://chaoyuesong.github.io/REACTO.

The perception of motion behavior in a dynamic environment holds significant importance for autonomous driving systems, wherein class-agnostic motion prediction methods directly predict the motion of the entire point cloud. While most existing methods rely on fully-supervised learning, the manual labeling of point cloud data is laborious and time-consuming. Therefore, several annotation-efficient methods have been proposed to address this challenge. Al-though effective, these methods rely on weak annotations or additional multi-modal data like images, and the potential benefits inherent in the point cloud sequence are still underexplored. To this end, we explore the feasibility of self-supervised motion prediction with only unlabeled Li-DAR point clouds. Initially, we employ an optimal transport solver to establish coarse correspondences between current and future point clouds as the coarse pseudo motion labels. Training models directly using such coarse labels leads to noticeable spatial and temporal prediction in-consistencies. To mitigate these issues, we introduce three simple spatial and temporal regularization losses, which fa-cilitate the self-supervised training process effectively. Experimental results demonstrate the significant superiority of our approach over the state-of-the-art self-supervised methods. Code will be available at https://github.com/kwwcv/SelfMotion.

Editable 3D-aware generation, which supports user-interacted editing, has witnessed rapid development re-cently. However, existing editable 3D GANs either fail to achieve high-accuracy local editing or suffer from huge computational costs. We propose AttriHuman-3D, an ed-itable 3D human generation model, which address the aforementioned problems with attribute decomposition and indexing. The core idea of the proposed model is to generate all attributes (e.g. human body, hair, clothes and so on) in an overall attribute space with six feature planes, which are then decomposed and manipulated with different attribute indexes. To precisely extract features of different attributes from the generated feature planes, we propose a novel at-tribute indexing method as well as an orthogonal projection regularization to enhance the disentanglement. We also introduce a hyper-latent training strategy and an attribute-specific sampling strategy to avoid style entanglement and misleading punishment from the discriminator. Our method allows users to interactively edit selected attributes in the generated 3D human avatars while keeping others fixed. Both qualitative and quantitative experiments demonstrate that our model provides a strong disentanglement between different attributes, allows fine-grained image editing and generates high-quality 3D human avatars.

Sarcasm is an intricate expression phenomenon and has garnered increasing attentions over the recent years, especially for multimodal contexts such as videos.Nevertheless, despite being a significant aspect of human sentiment, the effect of sarcasm is consistently overlooked in sentiment analysis.Videos with sarcasm often convey sentiments that diverge or even contradict their explicit messages.Prior works mainly concentrate on simply modeling sarcasm and sentiment features by utilizing the Multi-Task Learning (MTL) framework, which we found introduces detrimental interplays between the sarcasm detection task and sentiment analysis task.Therefore, this study explores the effective enhancement of video sentiment analysis through the incorporation of sarcasm information.To this end, we propose the Progressively Sentiment-oriented Sarcasm Refinement and Integration (PS2RI) framework, which focuses on modeling sentiment-oriented sarcasm features to enhance sentiment prediction.Instead of naively combining sarcasm detection and sentiment prediction under an MTL framework, PS2RI iteratively performs the sentiment-oriented sarcasm refinement and sarcasm integration operations within the sentiment recognition framework, in order to progressively learn sarcasm-aware sentiment feature without suffering the detrimental interplays caused by information irrelevant to the sentiment analysis task.Extensive experiments are conducted to validate the effectiveness of our approach.Code is available at https://github.com/tiggers23/PS2RI.

3D content creation has long been a complex and time-consuming process, often requiring specialized skills and resources. While re- cent advancements have allowed for text-guided 3D object and scene generation, they still fall short of providing sufficient control over the generation process, leading to a gap between the user’s creative vision and the generated results. In this paper, we present iControl3D, a novel interactive system that empowers users to gen- erate and render customizable 3D scenes with precise control. To this end, a 3D creator interface has been developed to provide users with fine-grained control over the creation process. Technically, we leverage 3D meshes as an intermediary proxy to iteratively merge individual 2D diffusion-generated images into a cohesive and uni- fied 3D scene representation. To ensure seamless integration of 3D meshes, we propose to perform boundary-aware depth alignment before fusing the newly generated mesh with the existing one in 3D space. Additionally, to effectively manage depth discrepancies between remote content and foreground, we propose to model re- mote content separately with an environment map instead of 3D meshes. Finally, our neural rendering interface enables users to build a radiance field of their scene online and navigate the entire scene. Extensive experiments have been conducted to demonstrate the effectiveness of our system. The code will be made available at https://github.com/xingyi- li/iControl3D.

Large pre-trained vision-language models like CLIP have shown amazing zero-shot recognition performance. To adapt pre-trained vision-language models to downstream tasks, recent studies have focused on the learnable context + class name paradigm, which learns continuous prompt contexts on downstream datasets. In practice, the learned prompt context tends to overfit the base categories and cannot generalize well to novel categories out of the training data. Recent works have also noticed this problem and have proposed several improvements. In this work, we draw a new insight based on empirical analysis, that is, uninformative class names lead to degraded base-to-novel generalization performance in prompt learning, which is usually overlooked by existing works. Under this motivation, we advocate to improve the base-to-novel generalization performance of prompt learning by enhancing the semantic richness of class names. We coin our approach as the Information Disengagement based Associative Prompt Learning (IDAPL) mechanism which considers the associative, meanwhile, decoupled learning of prompt context and class name embedding. IDAPL can effectively alleviate the phenomenon of learnable context overfitting to base classes, meanwhile, learning more informative semantic representation of base classes by fine-tuning the class name embedding, leading to improved performance on both base and novel classes. Experimental results on eleven widely used few-shot learning benchmarks clearly validate the effectiveness of our proposed approach. Code is available at https://github.com/tiggers23/IDAPL

Expressive Human Mesh Recovery (HMR) involves reconstructing the 3D human body, including hands and face, from RGB images. It is difficult because humans are highly deformable, and hands are small and frequently occluded. Recent approaches have attempted to mitigate these issues using large datasets and models, but these solutions remain imperfect. Specifically, whole-body estimation models often inaccurately estimate hand poses, while hand expert models struggle with severe occlusions. To overcome these limitations, we introduce a dual-path cross augmentation framework with a novel adaptation approach called HMR-Adapter that enhances existing large HMR models. HMR-Adapter significantly improves expressive HMR performance by injecting additional guidance from other body parts. This approach refines hand pose predictions by incorporating body pose information and uses additional hand features to enhance body pose estimation in whole-body models. Remarkably, an HMR-Adapter with about 30M parameters significantly improves expressive HMR results by combining the adapted large whole-body and hand expert models. We show extensive experiments and analysis to demonstrate the efficacy of our method.

Many few-shot segmentation (FSS) methods use cross attention to fuse support foreground (FG) into query features, regardless of the quadratic complexity. A recent advance Mamba can also well capture intra-sequence dependencies, yet the complexity is only linear. Hence, we aim to devise a cross (attention-like) Mamba to capture inter-sequence dependencies for FSS. A simple idea is to scan on support features to selectively compress them into the hidden state, which is then used as the initial hidden state to sequentially scan query features. Nevertheless, it suffers from (1) support forgetting issue: query features will also gradually be compressed when scanning on them, so the support features in hidden state keep reducing, and many query pixels cannot fuse sufficient support features; (2) intra-class gap issue: query FG is essentially more similar to itself rather than to support FG, i.e., query may prefer not to fuse support features but their own ones from the hidden state, yet the success of FSS relies on the effective use of support information. To tackle them, we design a hybrid Mamba network (HMNet), including (1) a support recapped Mamba to periodically recap the support features when scanning query, so the hidden state can always contain rich support information; (2) a query intercepted Mamba to forbid the mutual interactions among query pixels, and encourage them to fuse more support features from the hidden state. Consequently, the support information is better utilized, leading to better performance. Extensive experiments have been conducted on two public benchmarks, showing the superiority of HMNet. The code is available at https://github.com/Sam1224/HMNet.

Meta-learning offers a promising avenue for few-shot learning (FSL), enabling models to glean a generalizable feature embedding through episodic training on synthetic FSL tasks in a source domain. Yet, in practical scenarios where the target task diverges from that in the source domain, meta-learning based method is susceptible to over-fitting. To overcome this, we introduce a novel framework, Meta-Exploiting Frequency Prior for Cross-Domain Few-Shot Learning, which is crafted to comprehensively exploit the cross-domain transferable image prior that each image can be decomposed into complementary low-frequency content details and high-frequency robust structural characteristics. Motivated by this insight, we propose to decompose each query image into its high-frequency and low-frequency components, and parallel incorporate them into the feature embedding network to enhance the final category prediction. More importantly, we introduce a feature reconstruction prior and a prediction consistency prior to separately encourage the consistency of the intermediate feature as well as the final category prediction between the original query image and its decomposed frequency components. This allows for collectively guiding the network's meta-learning process with the aim of learning generalizable image feature embeddings, while not introducing any extra computational cost in the inference phase. Our framework establishes new state-of-the-art results on multiple cross-domain few-shot learning benchmarks.

In this article, we investigate self-supervised 3D scene flow estimation and class-agnostic motion prediction on point clouds. A realistic scene can be well modeled as a collection of rigidly moving parts, therefore its scene flow can be represented as a combination of rigid motion of these individual parts. Building upon this observation, we propose to generate pseudo scene flow labels for self-supervised learning through piecewise rigid motion estimation, in which the source point cloud is decomposed into local regions and each region is treated as rigid. By rigidly aligning each region with its potential counterpart in the target point cloud, we obtain a region-specific rigid transformation to generate its pseudo flow labels. To mitigate the impact of potential outliers on label generation, when solving the rigid registration for each region, we alternately perform three steps: establishing point correspondences, measuring the confidence for the correspondences, and updating the rigid transformation based on the correspondences and their confidence. As a result, confident correspondences will dominate label generation, and a validity mask will be derived for the generated pseudo labels. By using the pseudo labels together with their validity mask for supervision, models can be trained in a self-supervised manner. Extensive experiments on FlyingThings3D and KITTI datasets demonstrate that our method achieves new state-of-the-art performance in self-supervised scene flow learning, without any ground truth scene flow for supervision, even performing better than some supervised counterparts. Additionally, our method is further extended to class-agnostic motion prediction and significantly outperforms previous state-of-the-art self-supervised methods on nuScenes dataset.

Large-scale datasets with point-wise semantic and instance labels are crucial to 3D instance segmentation but also expensive. To leverage unlabeled data, previous semi-supervised 3D instance segmentation approaches have explored self-training frameworks, which rely on high-quality pseudo labels for consistency regularization. They intuitively utilize both instance and semantic pseudo labels in a joint learning manner. However, semantic pseudo labels contain numerous noise derived from the imbalanced category distribution and natural confusion of similar but distinct categories, which leads to severe collapses in self-training. Motivated by the observation that 3D instances are non-overlapping and spatially separable, we ask whether we can solely rely on instance consistency regularization for improved semi-supervised segmentation. To this end, we propose a novel self-training network InsTeacher3D to explore and exploit pure instance knowledge from unlabeled data. We first build a parallel base 3D instance segmentation model DKNet, which distinguishes each instance from the others via discriminative instance kernels without reliance on semantic segmentation. Based on DKNet, we further design a novel instance consistency regularization framework to generate and leverage high-quality instance pseudo labels. Experimental results on multiple large-scale datasets show that the InsTeacher3D significantly outperforms prior state-of-the-art semi-supervised approaches.

One-shot object detection (OSOD) without fine-tuning has recently garnered considerable attention and research focus. It aims to directly detect novel-class objects in the target image by providing merely one support image patch without undergoing the fine-tuning stage. However, most existing methods adopt image pair matching regardless of the scale inconsistency and spatial semantic mismatch of image pairs, which limits their ability to acquire high-quality target-support related features. This paper addresses these limitations by incorporating cross-scale contexts and semantic-consistent cues that are robust against the challenges of scarce and ambiguous matching. Specifically, we first introduce a simple yet effective Aggregation-Transformer-based Pyramid (ATP) module to explore the long-range cross-scale spatial interactions by employing the customized size-aware aggregation approach and the vanilla transformer encoder, thus the coarse-to-fine local image patterns are optimally utilized. Furthermore, we formulate the 4D contrastive cross-correlation tensor for instance-level features matching and suggest a Geometric Consistent Correlation (GCC) module that utilizes the bidirectional spatial-aware convolutions to extract the long-range semantic correspondences for target-support pairs. Additionally, a Channel Contrastive Learning (CCL) branch is adopted to complement the inter-channel interactions between target-support pairs for the GCC module. Extensive experiments demonstrate that our approach significantly outperforms the previous state-of-the-art methods by 6.5% and 2.1% on PASCAL VOC and COCO datasets for unseen classes, respectively.

Semantic segmentation on 3D point clouds is an important task for 3D scene understanding. While dense labeling on 3D data is expensive and time-consuming, only a few works address weakly supervised semantic point cloud segmentation methods to relieve the labeling cost by learning from simpler and cheaper labels. Meanwhile, there are still huge performance gaps between existing weakly supervised methods and state-of-the-art fully supervised methods. In this paper, we propose Dense Supervision Propagation (DSP) to train a semantic point cloud segmentation network with only a small portion of points being labeled. We argue that we can better utilize the limited supervision information as we densely propagate the supervision signal from the labeled points to other points within and across the input samples. Specifically, we propose a cross-sample feature reallocating module to transfer similar features and therefore re-route the gradients across two samples with common classes and an intra-sample feature redistribution module to propagate supervision signals on unlabeled points across and within point cloud samples. We conduct extensive experiments on public datasets S3DIS and ScanNet. Our weakly supervised method with only 10% and 1% of labels can produce competitive results with the fully supervised counterpart.

Indoor self-localization has become a highly desirable system function for smartphones. The existing systems based on imaging, radio frequency, and geomagnetic sensing may have sub-optimal performance when their limiting factors prevail. In this paper, we present a new indoor simultaneous localization and mapping (SLAM) system that is based on the smartphone's built-in audio hardware and inertial measurement unit (IMU). Our system uses a smartphone's loudspeaker to emit near-inaudible chirps and then the microphone to record the acoustic echoes from the indoor environment. The echoes contain the smartphone's location information with sub-meter granularity. To enable SLAM, we apply contrastive learning to train an echoic location feature (ELF) extractor, such that the loop closures on the smartphone's trajectory can be accurately detected from the associated ELF trace. The detection results effectively regulate the IMU-based trajectory reconstruction. The reconstructed trajectories are used fortrajectory map superimpositionandroom geometry reconstruction. Extensive experiments show that our SLAM achieves median localization errors of$\text{0.1}\,\text{m}$,$\text{0.53}\,\text{m}$, and$\text{0.4}\,\text{m}$in a living room, an office, and a shopping mall, and outperforms both the Wi-Fi and geomagnetic SLAM systems. The room geometry reconstruction achieves up to 4× lower errors compared with the latest echo-based approaches.

Few-shot point cloud classification is currently an under-explored problem which aims to learn a point cloud classifier for novel categories given a few annotated training data. Most existing methods achieve classification by matching a query point cloud to the most similar support category at the global representation level. However, due to the complicated structure of the point cloud and scarce available training data, the global representations of the point clouds and categories are of low quality, limiting the matching accuracy. Therefore, in this paper, we propose the Component-Aware Matching Network (CMNet) that matches the point clouds at the component level in addition to the global level. Specifically, we construct a component set for each query point cloud and support category and develop a metric to measure the similarity between two component sets. The final prediction is the weighted sum of the global and component matching probabilities. Besides, we carefully devise a component matching pretraining scheme for CMNet to enhance its ability to extract component features, further improving its performance. To evaluate the effectiveness of our design, we conduct comprehensive experiments on three benchmarks, namely ModelNet-FS, ShapeNet-FS and Shrec-FS. As a result, CMNet consistently outperforms the existing methods with significant margins in all the experiments of the three benchmarks and sets new state-of-the-art performance.

Humans tend to mine objects by learning from a group of images or several frames of video since we live in a dynamic world. In the computer vision area, many researchers focus on co-segmentation (CoS), co-saliency detection (CoSD) and video salient object detection (VSOD) to discover the co-occurrent objects. However, previous approaches design different networks for these similar tasks separately, and they are difficult to apply to each other. Besides, they fail to take full advantage of the cues among inter- and intra-feature within a group of images. In this paper, we introduce a unified framework to tackle these issues from a unified view, term asUFGS(UnifiedFramework forGroup-basedSegmentation). Specifically, we first introduce a transformer block, which views the image feature as a patch token and then captures their long-range dependencies through the self-attention mechanism. This can help the network to excavate the patch-structured similarities among the relevant objects. Furthermore, we propose an intra-MLP learning module to produce self-mask to enhance the network to avoid partial activation. Extensive experiments on four CoS benchmarks (PASCAL, iCoseg Internet and MSRC), three CoSD benchmarks (Cosal2015, CoSOD3k, and CocA) and five VSOD benchmarks (DAVIS$_{16}$, FBMS, ViSal, SegV2, and DAVSOD) show that our method outperforms other state-of-the-arts on three different tasks in both accuracy and speed by using the same network architecture, which can reach 140 FPS in real-time.

Depth information plays a pivotal role in numerous computer vision applications, including autonomous driving, 3D reconstruction, and 3D content generation. When deploying depth estimation models in practical applications, it is essential to ensure that the models have strong generalization capabilities. However, existing depth estimation methods primarily concentrate on robust single-image depth estimation, leading to the occurrence of flickering artifacts when applied to video inputs. On the other hand, video depth estimation methods either consume excessive computational resources or lack robustness. To address the above issues, we propose ViTA, a video transformer adaptor, to estimate temporally consistent video depth in the wild. In particular, we leverage a pre-trained image transformer (i.e., DPT) and introduce additional temporal embeddings in the transformer blocks. Such designs enable our ViTA to output reliable results given an unconstrained video. Besides, we present a spatio-temporal consistency loss for supervision. The spatial loss computes the per-pixel discrepancy between the prediction and the ground truth in space, while the temporal loss regularizes the inconsistent outputs of the same point in consecutive frames. To find the correspondences between consecutive frames, we design a bi-directional warping strategy based on the forward and backward optical flow. During inference, our ViTA no longer requires optical flow estimation, which enables it to estimate spatially accurate and temporally consistent video depth maps with fine-grained details in real time. We conduct a detailed ablation study to verify the effectiveness of the proposed components. Extensive experiments on the zero-shot cross-dataset evaluation demonstrate that the proposed method is superior to previous methods.

If we compare how humans reason and how deep models reason, humans reason in a symbolic manner with a formal language called logic, while most deep models reason in black-box. A natural question to ask is “Do the trained deep models reason similar as humans?” or “Can we explain the reasoning of deep models in the language of logic?”. In this work, we presentNeurLogXto explain the reasoning process of deep vision language models in the language of logic. Given a trained vision language model, our method starts by generating reasoning facts through augmenting the input data. We then develop a differentiable inductive logic programming framework to learn interpretable logic rules from the facts. We show our results on various popular vision language models. Interestingly, we observe that almost all of the tested models can reason logically.

Embodied question answering (EQA) is a recently emerged research field in which an agent is asked to answer the user's questions by exploring the environment and collecting visual information. Plenty of researchers turn their attention to the EQA field due to its broad potential application areas, such as in-home robots, self-driven mobile, and personal assistants. High-level visual tasks, such as EQA, are susceptible to noisy inputs, because they have complex reasoning processes. Before the profits of the EQA field can be applied to practical applications, good robustness against label noise needs to be equipped. To tackle this problem, we propose a novel label noise-robust learning algorithm for the EQA task. First, a joint training co-regularization noise-robust learning method is proposed for noisy filtering of the visual question answering (VQA) module, which trains two parallel network branches by one loss function. Then, a two-stage hierarchical robust learning algorithm is proposed to filter out noisy navigation labels in both trajectory level and action level. Finally, by taking purified labels as inputs, a joint robust learning mechanism is given to coordinate the work of the whole EQA system. Empirical results demonstrate that, under extremely noisy environments (45% of noisy labels) and low-level noisy environments (20% of noisy labels), the robustness of deep learning models trained by our algorithm is superior to the existing EQA models in noisy environments.

Understanding the motion behavior of dynamic environments is vital for autonomous driving, leading to increasing attention in class-agnostic motion prediction in LiDAR point clouds. Outdoor scenes can often be decomposed into mobile foregrounds and static backgrounds, which enables us to associate motion understanding with scene parsing. Based on this observation, we study a novel weakly supervised motion prediction paradigm, where fully or partially (1 %, 0.1%) annotated foreground/background binary masks are used for supervision, rather than using expensive motion annotations. To this end, we propose a two-stage weakly supervised approach, where the segmentation model trained with the incomplete binary masks in Stage1 will facilitate the self-supervised learning of the motion prediction network in Stage2 by estimating possible moving foregrounds in advance. Furthermore, for robust self-supervised motion learning, we design a Consistency-aware Chamfer Distance loss by exploiting multi-frame information and explicitly suppressing potential outliers. Comprehensive experiments show that, with fully or partially binary masks as supervision, our weakly supervised models surpass the self-supervised models by a large margin and perform on par with some supervised ones. This further demonstrates that our approach achieves a good compromise between annotation effort and performance.

We present 3D Cinemagraphy, a new technique that mar-ries 2D image animation with 3D photography. Given a single still image as input, our goal is to generate a video that contains both visual content animation and camera motion. We empirically find that naively combining existing 2D image animation and 3D photography methods leads to obvious artifacts or inconsistent animation. Our key insight is that representing and animating the scene in 3D space offers a natural solution to this task. To this end, we first convert the input image into feature-based layered depth images using predicted depth values, followed by unprojecting them to a feature point cloud. To animate the scene, we perform motion estimation and lift the 2D motion into the 3D scene flow. Finally, to resolve the problem of hole emer-gence as points move forward, we propose to bidirectionally displace the point cloud as per the scene flow and synthe-size novel views by separately projecting them into target image planes and blending the results. Extensive experiments demonstrate the effectiveness of our method. A user study is also conducted to validate the compelling rendering results of our method.

In this paper, we investigate an open research task of generating controllable 3D textured shapes from the given textual descriptions. Previous works either require ground truth caption labeling or extensive optimization time. To resolve these issues, we present a novel framework, TAPS3D, to train a text-guided 3D shape generator with pseudo captions. Specifically, based on rendered 2D images, we retrieve relevant words from the CLIP vocabulary and construct pseudo captions using templates. Our constructed captions provide high-level semantic supervision for generated 3D shapes. Further, in order to produce fine-grained textures and increase geometry diversity, we propose to adopt low-level image regularization to enable fake-rendered images to align with the real ones. During the inference phase, our proposed model can generate 3D textured shapes from the given text without any additional optimization. We conduct extensive experiments to analyze each of our proposed components and show the efficacy of our framework in generating high-fidelity 3D textured and text-relevant shapes. Code is available at https://github.com/plusmultiply/TAPS3D

Deep neural networks (DNNs) are widely applied for nowadays 3D surface reconstruction tasks and such methods can be further divided into two categories, which respectively warp templates explicitly by moving vertices or represent 3D surfaces implicitly as signed or unsigned distance functions. Taking advantage of both advanced explicit learning process and powerful representation ability of implicit functions, we propose a novel 3D representation method, Neural Vector Fields (NVF). It not only adopts the explicit learning process to manipulate meshes directly, but also leverages the implicit representation of unsigned distance functions (UDFs) to break the barriers in resolution and topology. Specifically, our method first predicts the displacements from queries towards the surface and models the shapes as Vector Fields. Rather than relying on network differentiation to obtain direction fields as most existing UDF-based methods, the produced vector fields encode the distance and direction fields both and mitigate the ambiguity at “ridge” points, such that the calculation of direction fields is straightforward and differentiation-free. The differentiation-free characteristic enables us to further learn a shape codebook via Vector Quantization, which encodes the cross-object priors, accelerates the training procedure, and boosts model generalization on cross-category reconstruction. The extensive experiments on surface reconstruction benchmarks indicate that our method outperforms those state-of-the-art methods in different evaluation scenarios including watertight vs non-watertight shapes, category-specific vs category-agnostic reconstruction, category-unseen reconstruction, and cross-domain reconstruction. Our code is released at https://github.com/Wi-sc/NVF.

Continual semantic segmentation (CSS) aims to extend an existing model to tackle unseen tasks while retaining its old knowledge. Naively fine-tuning the old model on new data leads to catastrophic forgetting. A common solution is knowledge distillation (KD), where the output distribution of the new model is regularized to be similar to that of the old model. However, in CSS, this is challenging because of the background shift issue. Existing KD-based CSS methods continue to suffer from confusion between the background and novel classes since they fail to establish a reliable class correspondence for distillation. To address this issue, we propose a new label-guided knowledge distillation (LGKD) loss, where the old model output is expanded and transplanted (with the guidance of the ground truth label) to form a semantically appropriate class correspondence with the new model output. Consequently, the useful knowledge from the old model can be effectively distilled into the new model without causing confusion. We conduct extensive experiments on two prevailing CSS benchmarks, Pascal-VOC and ADE20K, where our LGKD significantly boosts the performance of three competing methods, especially on novel mIoU by up to +76%, setting new state-of-the-art. Finally, to further demonstrate its generalization ability, we introduce the first CSS benchmark for 3D point cloud based on ScanNet, along with several re-implemented baselines for comparison. Experiments show that LGKD is versatile in both 2D and 3D modalities without requiring ad hoc design. Codes are available at https://github.com/Ze-Yang/LGKD.

Video depth estimation aims to infer temporally consistent depth. Some methods achieve temporal consistency by finetuning a single-image depth model during test time using geometry and re-projection constraints, which is inefficient and not robust. An alternative approach is to learn how to enforce temporal consistency from data, but this requires well-designed models and sufficient video depth data. To address these challenges, we propose a plug-and-play framework called Neural Video Depth Stabilizer (NVDS) that stabilizes inconsistent depth estimations and can be applied to different single-image depth models without extra effort. We also introduce a large-scale dataset, Video Depth in the Wild (VDW), which consists of 14,203 videos with over two million frames, making it the largest natural-scene video depth dataset to our knowledge. We evaluate our method on the VDW dataset as well as two public benchmarks and demonstrate significant improvements in consistency, accuracy, and efficiency compared to previous approaches. Our work serves as a solid baseline and provides a data foundation for learning-based video depth models. We will release our dataset and code for future research.

Instance segmentation on 3D point clouds has been attracting increasing attention due to its wide applications, especially in scene understanding areas. However, most existing methods operate on fully annotated data while manually preparing ground-truth labels at point-level is very cumbersome and labor-intensive. To address this issue, we propose a novel weakly supervised method RWSeg that only requires labeling one object with one point. With these sparse weak labels, we introduce a unified framework with two branches to propagate semantic and instance information respectively to unknown regions using self-attention and a cross-graph random walk method. Specifically, we propose a Cross-graph Competing Random Walks (CRW) algorithm that encourages competition among different instance graphs to resolve ambiguities in closely placed objects, improving instance assignment accuracy. RWSeg generates high-quality instance-level pseudo labels. Experimental results on ScanNet-v2 and S3DIS datasets show that our approach achieves comparable performance with fully-supervised methods and outperforms previous weakly-supervised methods by a substantial margin.

The key to the success of few-shot segmentation (FSS) lies in how to effectively utilize support samples. Most solutions compress support foreground (FG) features into prototypes, but lose some spatial details. Instead, others use cross attention to fuse query features with uncompressed support FG. Query FG could be fused with support FG, however, query background (BG) cannot find matched BG features in support FG, yet inevitably integrates dissimilar features. Besides, as both query FG and BG are combined with support FG, they get entangled, thereby leading to ineffective segmentation. To cope with these issues, we design a self-calibrated cross attention (SCCA) block. For efficient patch-based attention, query and support features are firstly split into patches. Then, we design a patch alignment module to align each query patch with its most similar support patch for better cross attention. Specifically, SCCA takes a query patch as Q, and groups the patches from the same query image and the aligned patches from the support image as K&V. In this way, the query BG features are fused with matched BG features (from query patches), and thus the aforementioned issues will be mitigated. Moreover, when calculating SCCA, we design a scaled-cosine mechanism to better utilize the support features for similarity calculation. Extensive experiments conducted on PASCAL-5iand COCO-20idemonstrate the superiority of our model, e.g., the mIoU score under 5-shot setting on COCO-20iis 5.6%+ better than previous state-of-the-arts. The code is available at https://github.com/Sam1224/SCCAN.

Multi-source domain adaptation (MSDA) aims to leverage the knowledge in multiple source domains to assist the prediction in a target domain, where the source and target domains have different data distributions. This paper presents a MSDA model to investigate both domain discrepancy and domain relevance, whose interactions are also exploited to gradually refine the learning performance. Particularly, the proposed model contains two components, i.e., feature spaces learning and transferred weights learning. The former one minimizes the domain discrepancy and the latter one evaluates the domain relevance. Experimental results on several real-world datasets demonstrate the effectiveness of the proposed model.

We study the problem of synthesizing a long-term dynamic video from only a single image. This is challenging since it requires consistent visual content movements given large camera motions. Existing methods either hallucinate inconsistent perpetual views or struggle with long camera trajectories. To address these issues, it is essential to estimate the underlying 4D (including 3D geometry and scene motion) and fill in the occluded regions. To this end, we present Make-It-4D, a novel method that can generate a consistent long-term dynamic video from a single image. On the one hand, we utilize layered depth images (LDIs) to represent a scene, and they are then unprojected to form a feature point cloud. To animate the visual content, the feature point cloud is displaced based on the scene flow derived from motion estimation and the corresponding camera pose. Such 4D representation enables our method to maintain the global consistency of the generated dynamic video. On the other hand, we fill in the occluded regions by using a pre-trained diffusion model to inpaint and outpaint the input image. This enables our method to work under large camera motions. Benefiting from our design, our method can be training-free which saves a significant amount of training time. Experimental results demonstrate the effectiveness of our approach, which showcases compelling rendering results.

The current state of re-identification (ReID) models poses limitations to their applicability in the open world, as they are primarily designed and trained for specific categories like person or vehicle. In light of the importance of ReID technology for tracking wildlife populations and migration patterns, we propose a new task called ``Re-identify Any Animal in the Wild'' (ReID-AW). This task aims to develop a ReID model capable of handling any unseen wildlife category it encounters. To address this challenge, we have created a comprehensive dataset called Wildlife-71, which includes ReID data from 71 different wildlife categories. This dataset is the first of its kind to encompass multiple object categories in the realm of ReID. Furthermore, we have developed a universal re-identification model named UniReID specifically for the ReID-AW task. To enhance the model's adaptability to the target category, we employ a dynamic prompting mechanism using category-specific visual prompts. These prompts are generated based on knowledge gained from a set of pre-selected images within the target category. Additionally, we leverage explicit semantic knowledge derived from the large-scale pre-trained language model, GPT-4. This allows UniReID to focus on regions that are particularly useful for distinguishing individuals within the target category. Extensive experiments have demonstrated the remarkable generalization capability of our UniReID model. It showcases promising performance in handling arbitrary wildlife categories, offering significant advancements in the field of ReID for wildlife conservation and research purposes.

Embodied Question Answering (EQA) is a newly defined research area where an agent is required to answer the user's questions by exploring the real-world environment. It has attracted increasing research interests due to its broad applications in personal assistants and in-home robots. Most of the existing methods perform poorly in terms of answering and navigation accuracy due to the absence of fine-level semantic information, stability to the ambiguity, and 3D spatial information of the virtual environment. To tackle these problems, we propose a depth and segmentation based visual attention mechanism for Embodied Question Answering. First, we extract local semantic features by introducing a novel high-speed video segmentation framework. Then guided by the extracted semantic features, a depth and segmentation based visual attention mechanism is proposed for the Visual Question Answering (VQA) sub-task. Further, a feature fusion strategy is designed to guide the navigator's training process without much additional computational cost. The ablation experiments show that our method effectively boosts the performance of the VQA module and navigation module, leading to 4.9 % and 5.6 % overall improvement in EQA accuracy on House3D and Matterport3D datasets respectively.

The goal of 3D pose transfer is to transfer the pose from the source mesh to the target mesh while preserving the identity information (e.g., face, body shape) of the target mesh. Deep learning-based methods improved the efficiency and performance of 3D pose transfer. However, most of them are trained under the supervision of the ground truth, whose availability is limited in real-world scenarios. In this work, we present X-DualNet, a simple yet effective approach that enables unsupervised 3D pose transfer. In X-DualNet, we introduce a generator G which contains correspondence learning and pose transfer modules to achieve 3D pose transfer. We learn the shape correspondence by solving an optimal transport problem without any key point annotations and generate high-quality meshes with our elastic instance normalization (ElaIN) in the pose transfer module. With G as the basic component, we propose a cross consistency learning scheme and a dual reconstruction objective to learn the pose transfer without supervision. Besides that, we also adopt an as-rigid-as-possible deformer in the training process to fine-tune the body shape of the generated results. Extensive experiments on human and animal data demonstrate that our framework can successfully achieve comparable performance as the state-of-the-art supervised approaches.

Food is significant to human daily life. In this paper, we are interested in learning structural representations for lengthy recipes, that can benefit the recipe generation and food cross-modal retrieval tasks. Different from the common vision-language data, here the food images contain mixed ingredients and target recipes are lengthy paragraphs, where we do not have annotations on structure information. To address the above limitations, we propose a novel method to unsupervisedly learn the sentence-level tree structures for the cooking recipes. Our approach brings together several novel ideas in a systematic framework: (1) exploiting an unsupervised learning approach to obtain the sentence-level tree structure labels before training; (2) generating trees of target recipes from images with the supervision of tree structure labels learned from (1); and (3) integrating the learned tree structures into the recipe generation and food cross-modal retrieval procedure. Our proposed model can produce good-quality sentence-level tree structures and coherent recipes. We achieve the state-of-the-art recipe generation and food cross-modal retrieval performance on the benchmark Recipe1M dataset.

In this work, we develop methods for few-shot image classification from a new perspective of optimal matching between image regions. We employ the Earth Mover's Distance (EMD) as a metric to compute a structural distance between dense image representations to determine image relevance. The EMD generates the optimal matching flows between structural elements that have the minimum matching cost, which is used to calculate the image distance for classification. To generate the important weights of elements in the EMD formulation, we design a cross-reference mechanism, which can effectively alleviate the adverse impact caused by the cluttered background and large intra-class appearance variations. To implement k-shot classification, we propose to learn a structured fully connected layer that can directly classify dense image representations with the EMD. Based on the implicit function theorem, the EMD can be inserted as a layer into the network for end-to-end training. Our extensive experiments validate the effectiveness of our algorithm which outperforms state-of-the-art methods by a significant margin on five widely used few-shot classification benchmarks, namely, miniImageNet, tieredImageNet, Fewshot-CIFAR100 (FC100), Caltech-UCSD Birds-200-2011 (CUB), and CIFAR-FewShot (CIFAR-FS). We also demonstrate the effectiveness of our method on the image retrieval task in our experiments.

In real-world LiDAR-based applications, data is generated in the form of 3D point cloud sequences or 4D point clouds. However, the topic of semantic segmentation on 4D point clouds is under-investigated and existing methods are still not able to achieve satisfactory performance to meet the requirement for real-world applications. The temporal information across different point clouds plays an important role in dynamic scene understanding, which is not well explored in existing work. In this paper, we focus on exploring effective temporal information across two consecutive point clouds for semantic segmentation on point cloud sequences. To this end, we design three novel modules to enhance the features of target frames by extracting different temporal information in the local regions and global regions. Experimental results on SemanticKITTI and SemanticPOSS demonstrate that our method achieves superior performance in 4D semantic segmentation by utilizing temporal information.

Natural language BERTs are trained with language corpus in a self-supervised manner. Unlike natural language BERTs, vision language BERTs need paired data to train, which restricts the scale of VL-BERT pretraining. We propose a self-training approach that allows training VL-BERTs from unlabeled image data. The proposed method starts with our unified conditional model– a vision language BERT model that can perform zero-shot conditional generation. Given different conditions, the unified conditional model can generate captions, dense captions, and even questions. We use the labeled image data to train a teacher model and use the trained model to generate pseudo captions on unlabeled image data. We then combine the labeled data and pseudo labeled data to train a student model. The process is iterated by putting the student model as a new teacher. By using the proposed self-training approach and only 300k unlabeled extra data, we are able to get competitive or even better performances compared to the models of similar model size trained with 3 million extra image data.

Single-view 3D object reconstruction is a fundamental and challenging computer vision task that aims at recovering 3D shapes from single-view RGB images. Most existing deep learning based reconstruction methods are trained and evaluated on the same categories, and they cannot work well when handling objects from novel categories that are not seen during training. Focusing on this issue, this paper tackles Single-view 3D Mesh Reconstruction, to study the model generalization on unseen categories and encourage models to reconstruct objects literally. Specifically, we propose an end-to-end two-stage network, GenMesh, to break the category boundaries in reconstruction. Firstly, we factorize the complicated image-to-mesh mapping into two simpler mappings, i.e., image-to-point mapping and point-to-mesh mapping, while the latter is mainly a geometric problem and less dependent on object categories. Secondly, we devise a local feature sampling strategy in 2D and 3D feature spaces to capture the local geometry shared across objects to enhance model generalization. Thirdly, apart from the traditional point-to-point supervision, we introduce a multi-view silhouette loss to supervise the surface generation process, which provides additional regularization and further relieves the overfitting problem. The experimental results show that our method significantly outperforms the existing works on the ShapeNet and Pix3D under different scenarios and various metrics, especially for novel objects.

Few-shot object detection (FSOD), which aims at learning a generic detector that can adapt to unseen tasks with scarce training samples, has witnessed consistent improvement recently. However, most existing methods ignore the efficiency issues, e.g., high computational complexity and slow adaptation speed. Notably, efficiency has become an increasingly important evaluation metric for few-shot techniques due to an emerging trend toward embedded AI. To this end, we present an efficient pretrain-transfer framework (PTF) baseline with no computational increment, which achieves comparable results with previous state-of-the-art (SOTA) methods. Upon this baseline, we devise an initializer named knowledge inheritance (KI) to reliably initialize the novel weights for the box classifier, which effectively facilitates the knowledge transfer process and boosts the adaptation speed. Within the KI initializer, we propose an adaptive length re-scaling (ALR) strategy to alleviate the vector length inconsistency between the predicted novel weights and the pretrained base weights. Finally, our approach not only achieves the SOTA results across three public benchmarks, i.e., PASCAL VOC, COCO and LVIS, but also exhibits high efficiency with $1.8-100\times $ faster adaptation speed against the other methods on COCO/LVIS benchmark during few-shot transfer. To our best knowledge, this is the first work to consider the efficiency problem in FSOD. We hope to motivate a trend toward powerful yet efficient few-shot technique development. The codes are publicly available at https://github.com/Ze-Yang/Efficient-FSOD.

Unsupervised domain adaptation has limitations when encountering label discrepancy between the source and target domains. While open-set domain adaptation approaches can address situations when the target domain has additional categories, these methods can only detect them but not further classify them. In this paper, we focus on a more challenging setting dubbed Domain Adaptive Zero-Shot Learning (DAZSL), which uses semantic embeddings of class tags as the bridge between seen and unseen classes to learn the classifier for recognizing all categories in the target domain when only the supervision of seen categories in the source domain is available. The main challenge of DAZSL is to perform knowledge transfer across categories and domain styles simultaneously. To this end, we propose a novel end-to-end learning mechanism dubbed Three-way Semantic Consistent Embedding (TSCE) to embed the source domain, target domain, and semantic space into a shared space. Specifically, TSCE learns domain-irrelevant categorical prototypes from the semantic embedding of class tags and uses them as the pivots of the shared space. The source domain features are aligned with the prototypes via their supervised information. On the other hand, the mutual information maximization mechanism is introduced to push the target domain features and prototypes towards each other. By this way, our approach can align domain differences between source and target images, as well as promote knowledge transfer towards unseen classes. Moreover, as there is no supervision in the target domain, the shared space may suffer from the catastrophic forgetting problem. Hence, we further propose a ranking-based embedding alignment mechanism to maintain the consistency between the semantic space and the shared space. Experimental results on both I2AwA and I2WebV clearly validate the effectiveness of our method. Code is available at https://github.com/tiggers23/TSCE-Domain-Adaptive-Zero-Shot-Learning.

Weakly supervised image segmentation trained with image-level labels usually suffers from inaccurate coverage of object areas during the generation of the pseudo groundtruth. This is because the object activation maps are trained with the classification objective and lack the ability to generalize. To improve the generality of the object activation maps, we propose a region prototypical network (RPNet) to explore the cross-image object diversity of the training set. Similar object parts across images are identified via region feature comparison. Object confidence is propagated between regions to discover new object areas while background regions are suppressed. Experiments show that the proposed method generates more complete and accurate pseudo object masks while achieving state-of-the-art performance on PASCAL VOC 2012 and MS COCO. In addition, we investigate the robustness of the proposed method on reduced training sets. The code is available athttps://github.com/liuweide01/RPNet-Weakly-Supervised-Segmentation.

We tackle the challenging task of few-shot segmentation in this work. It is essential for few-shot semantic segmentation to fully utilize the support information. Previous methods typically adopt masked average pooling over the support feature to extract the support clues as a global vector, usually dominated by the salient part and lost certain essential clues. In this work, we argue that every support pixel’s information is desired to be transferred to all query pixels and propose a Correspondence Matching Network (CMNet) with an Optimal Transport Matching module to mine out the correspondence between the query and support images. Besides, it is critical to fully utilize both local and global information from the annotated support images. To this end, we propose a Message Flow module to propagate the message along the inner-flow inside the same image and cross-flow between support and query images, which greatly helps enhance the local feature representations. Experiments on PASCAL VOC 2012, MS COCO, and FSS-1000 datasets show that our network achieves new state-of-the-art few-shot segmentation performance.

Current vision language pretraining models are dominated by methods using region visual features extracted from object detectors. Given their good performance, the extract-then-process pipeline significantly restricts the inference speed and therefore limits their real-world use cases. However, training vision language models from raw image pixels is difficult, as the raw image pixels give much less prior knowledge than region features. In this paper, we systematically study how to leverage auxiliary visual pretraining tasks to help training end-to-end vision language models. We introduce three types of visual losses that enable much faster convergence and better finetuning accuracy. Compared with region feature models, our end-to-end models could achieve similar or better performance on down-stream tasks and run more than 10 times faster during inference. Compared with other end-to-end models, our proposed method could achieve similar or better performance when pretrained for only 10% of the pretraining GPU hours.

The lack of sufficient training data has been one obstacle to fine-grained visual classification research because labeling subcategories generally requires specialist knowledge. As one optional approach to alleviating the data-hunger problem, leveraging web images as training data is drawing increasing attention. Nevertheless, web images potentially have false labels, which can misguide the training process. Although several works have been proposed to deal with label noise, it still can be difficult for the network to tackle complex real-world noisy labels without any prior knowledge. In the literature, we propose to leverage a small and clean meta-set to provide reliable prior knowledge for tackling noisy web images. Specifically, our method trains a network with two peer predicting heads, which learn from noisy web images (web head) and meta ones (meta head), respectively. The meta head produces pseudo soft labels for web images to revise their training loss, which can overcome the high noise ratio problem. Furthermore, a selection net is trained in a meta-learning strategy to identify in- and out-of-distribution noisy images. Then in-distribution ones are reused for training with pseudo soft labels produced by the meta head as supervision, while out-of-distribution ones are discarded. In this manner, the misguidance caused by label noise is remarkably alleviated and in-distribution noisy samples are properly exploited to boost model performance. The superiority of our proposed approach is demonstrated by mathematical theory with great interpretability as well as extensive experimental results on the real-world dataset WebFG-496.

3D point cloud analysis has drawn a lot of research attention due to its wide applications. However, collecting massive labelled 3D point cloud data is both time-consuming and labor-intensive. This calls for data-efficient learning methods. In this work we propose PointDisc, a point discriminative learning method to leverage self-supervisions for data-efficient 3D point cloud classification and segmentation. PointDisc imposes a novel point discrimination loss on the middle and global level features produced by the backbone network. This point discrimination loss enforces learned features to be consistent with points belonging to the corresponding local shape region and inconsistent with randomly sampled noisy points. We conduct extensive experiments on 3D object classification, 3D semantic and part segmentation, showing the benefits of PointDisc for data-efficient learning. Detailed analysis demonstrate that PointDisc learns unsupervised features that well capture local and global geometry.

Object localization aims to generate a tight bounding box for the target object, which is a challenging problem that has been deeply studied in recent years. Since collecting bounding-box labels is time-consuming and laborious, many researchers focus on weakly supervised object localization (WSOL). As the recent appealing self-supervised learning technique shows its powerful function in visual tasks, in this paper, we take the early attempt to explore unsupervised object localization by self-supervision. Specifically, we adopt different geometric transformations to image and utilize their parameters as pseudo labels for self-supervised learning. Then, the class-agnostic activation map (CAAM) is used to highlight the target object potential regions. However, such attention maps merely focus on the most discriminative part of the objects, which will affect the quality of the predicted bounding box. Based on the motivation that the activation maps of different transformations of the same image should be equivariant, we further design a siamese network that encodes the paired images and propose a joint graph cluster partition mechanism in an unsupervised manner to enhance the object co-occurrent regions. To validate the effectiveness of the proposed method, extensive experiments are conducted on CUB-200-2011, Stanford Cars and FGVC-Aircraft datasets. Experimental results show that our method outperforms state-of-the-art methods using the same level of supervision, even outperforms some weakly-supervised methods.

Existing point cloud segmentation methods require a large amount of annotated data, especially for the outdoor point cloud scene. Due to the complexity of the outdoor 3D scenes, manual annotations on the outdoor point cloud scene are time-consuming and expensive. In this paper, we study how to achieve scene understanding with limited annotated data. Treating 100 consecutive frames as a sequence, we divide the whole dataset into a series of sequences and annotate only 0.1% points in the first frame of each sequence to reduce the annotation requirements. This leads to a total annotation budget of 0.001%. We propose a novel temporal-spatial framework for effective weakly supervised learning to generate high-quality pseudo labels from these limited annotated data. Specifically, the frame-work contains two modules: an matching module in temporal dimension to propagate pseudo labels across different frames, and a graph propagation module in spatial dimension to propagate the information of pseudo labels to the entire point clouds in each frame. With only 0.001% annotations for training, experimental results on both SemanticKITTI and SemanticPOSS shows our weakly supervised two-stage framework is comparable to some existing fully supervised methods. We also evaluate our framework with 0.005% initial annotations on SemanticKITTI, and achieve a result close to fully supervised backbone model.

In this work, we focus on scene flow learning on point clouds in a self-supervised manner. A real-world scene can be well modeled as a collection of rigidly moving parts, therefore its scene flow can be represented as a combination of rigid motion of each part. Inspired by this observation, we propose to generate pseudo scene flow for self-supervised learning based on piecewise rigid motion estimation, in which the source point cloud is decomposed into a set of local regions and each region is treated as rigid. By rigidly aligning each region with its potential counterpart in the target point cloud, we obtain a region-specific rigid transformation to represent the flow, which together constitutes the pseudo scene flow labels of the entire scene to enable network training. Compared with most existing approaches relying on point-wise similarities for scene flow approximation, our method explicitly enforces region-wise rigid alignments, yielding locally rigid pseudo scene flow labels. We demonstrate the effectiveness of our self-supervised learning method on FlyingThings3D and KITTI datasets. Comprehensive experiments show that our method achieves new state-of-the-art performance in self-supervised scene flow learning, without any ground truth scene flow for supervision, even outperforming some super-vised counterparts.

Fine-tuning pre-trained models for downstream tasks is mainstream in deep learning. However, the pre-trained models are limited to be fine-tuned by data from a specific modality. For example, as a visual model, DenseNet cannot directly take the textual data as its input. Hence, although the large pre-trained models such as DenseNet or BERT have a great potential for the downstream recognition tasks, they have weaknesses in leveraging multimodal information, which is a new trend of deep learning. This work focuses on fine-tuning pre-trained unimodal models with multimodal inputs of image-text pairs and expanding them for image-text multimodal recognition. To this end, we propose the Multimodal Information Injection Plug-in (MI2P) which is attached to different layers of the unimodal models (e.g., DenseNet and BERT). The proposed MI2P unit provides the path to integrate the information of other modalities into the unimodal models. Specifically, MI2P performs cross-modal feature transformation by learning the fine-grained correlations between the visual and textual features. Through the proposed MI2P unit, we can inject the language information into the vision backbone by attending the word-wise textual features to different visual channels, as well as inject the visual information into the language backbone by attending the channel-wise visual features to different textual words. Armed with the MI2P attachments, the pre-trained unimodal models can be expanded to process multimodal data without the need to change the network structures.

In this paper, we propose to solve the problem from two novel aspects. First, instead of learning the decision boundaries between seen classes, as is done in standard close-set classification, we reserve space for unseen classes, such that images located in these areas are recognized as the unseen classes. Second, to effectively learn such decision boundaries, we propose to utilize the background features from seen classes. As these background regions do not significantly contribute to the decision of close-set classification, it is natural to use them as pseudo unseen classes for classifier learning. Our extensive experiments show that our proposed method not only outperforms multiple baselines but also sets new state-of-the-art results on three popular benchmarks, namely tieredImageNet, miniImageNet, and Caltech-USCD Birds-200-2011 (CUB).

This paper investigates an open research problem of generating text-image pairs to improve the training of fine-grained image-to-text cross-modal retrieval task, and proposes a novel framework for paired data augmentation by uncovering the hidden semantic information of StyleGAN2 model. Specifically, we first train a StyleGAN2 model on the given dataset. We then project the real images back to the latent space of StyleGAN2 to obtain the latent codes. To make the generated images manipulatable, we further introduce a latent space alignment module to learn the alignment between StyleGAN2 latent codes and the corresponding textual caption features. When we do online paired data augmentation, we first generate augmented text through random token replacement, then pass the augmented text into the latent space alignment module to output the latent codes, which are finally fed to StyleGAN2 to generate the augmented images. We evaluate the efficacy of our augmented data approach on two public cross-modal retrieval datasets, in which the promising experimental results demonstrate the augmented text-image pair data can be trained together with the original data to boost the image-to-text cross-modal retrieval performance.

We present three novel strategies to incorporate a parametric body model into a pixel-aligned implicit model for single-view clothed human reconstruction. Firstly, we introduce ray-based sampling, a novel technique that transforms a parametric model into a set of highly informative, pixel-aligned 2D feature maps. Next, we propose a new type of feature based on blendweights. Blendweight-based labels serve as soft human parsing labels and help to improve the structural fidelity of reconstructed meshes. Finally, we show how we can extract and capitalize on body part orientation information from a parametric model to further improve reconstruction quality. Together, these three techniques form our S-PIFu framework, which significantly outperforms state-of-the-arts methods in all metrics. Our code is available at https://github.com/kcyt/SPIFu.

Indoor self-localization is a highly demanded system function for smartphones. The current solutions based on inertial, radio frequency, and geomagnetic sensing may have degraded performance when their limiting factors take effect. In this paper, we present a new indoor simultaneous localization and mapping (SLAM) system that utilizes the smartphone's built-in audio hardware and inertial measurement unit (IMU). Our system uses a smartphone's loud-speaker to emit near-inaudible chirps and then the microphone to record the acoustic echoes from the indoor environment. Our profiling measurements show that the echoes carry location information with sub-meter granularity. To enable SLAM, we apply contrastive learning to construct an echoic location feature (ELF) extractor, such that the loop closures on the smartphone's trajectory can be accurately detected from the associated ELF trace. The detection results effectively regulate the IMU-based trajectory reconstruction. Extensive experiments show that our ELF-based SLAM achieves median localization errors of 0.1 m, 0.53 m, and 0.4m on the reconstructed trajectories in a living room, an office, and a shopping mall, and outperforms the Wi-Fi and geomagnetic SLAM systems.

To reduce the manpower consumption on box-level annotations, many weakly supervised object detection methods which only require image-level annotations, have been proposed recently. The training process in these methods is formulated into two steps. They firstly train a neural network under weak supervision to generate pseudo ground truths (PGTs). Then, these PGTs are used to train another network under full supervision. Compared with fully supervised methods, the training process in weakly supervised methods becomes more complex and time-consuming. Furthermore, overwhelming negative proposals are involved at the first step. This is neglected by most methods, which makes the training network biased towards to negative proposals and thus degrades the quality of the PGTs, limiting the training network performance at the second step. Online proposal sampling is an intuitive solution to these issues. However, lacking of adequate labeling, a simple online proposal sampling may make the training network stuck into local minima . To solve this problem, we propose an O nline Active P roposal Set G eneration (OPG) algorithm. Our OPG algorithm consists of two parts: Dynamic Proposal Constraint (DPC) and Proposal Partition (PP). DPC is proposed to dynamically determine different proposal sampling strategies according to the current training state. PP is used to score each proposal, part proposals into different sets and generate an active proposal set for the network optimization. Through experiments, our proposed OPG shows consistent and significant improvement on both datasets PASCAL VOC 2007 and 2012, yielding comparable performance to the state-of-the-art results.

Most existing action recognition approaches directly leverage the video-level features to recognize human actions from videos. Although these methods have made remarkable progress, the accuracy is still unsatisfied. When the test video involves complex backgrounds and activities, existing methods usually suffer from a significant drop in accuracy. Human action is inherently a high-level concept. Merely applying a video classification model without a detailed semantic understanding of the video content, e.g., objects, scene context, object motions, object interactions, is inadequate to tackle the challenges for action recognition. Fine-level semantic understanding of videos generates elementary semantic concepts from the raw video data, such as the semantics of objects and background regions. It can be employed to bridge the gap between the raw video data and the high-level concept of human actions. In this work, we leverage dense semantic segmentation masks, which encode rich semantic details, provide extra information for the network training, and improve the performance of action recognition. We propose a novel deep architecture which is named as Dense Semantics-Assisted Convolutional Neural Networks (DSA-CNNs) to effectively utilize dense semantic information of video by a bottom-up attention way in the spatial stream, while by the way of branch fusion in the temporal stream. To verify the effectiveness of our approach, we conduct extensive experiments on publicly available datasets – UCF101, HMDB51, and Kinetics. The experimental results demonstrate that our approach substantially improves existing methods and achieves very competitive performance. It also shows that our approach is superior to other related methods that utilize extra information for action recognition.

Video captioning targets interpreting the complex visual contents as text descriptions, which requires the model to fully understand video scenes including objects and their interactions. Prevailing methods adopt off-the-shelf object detection networks to give object proposals and use the attention mechanism to model the relations between objects. They often miss some undefined semantic concepts of the pretrained model and fail to identify exact predicate relationships between objects. In this paper, we investigate an open research task of generating text descriptions for the given videos, and propose Cross-Modal Graph (CMG) with meta concepts for video captioning. Specifically, to cover the useful semantic concepts in video captions, we weakly learn the corresponding visual regions for text descriptions, where the associated visual regions and textual words are named cross-modal meta concepts. We further build meta concept graphs dynamically with the learned cross-modal meta concepts. We also construct holistic video-level and local frame-level video graphs with the predicted predicates to model video sequence structures. We validate the efficacy of our proposed techniques with extensive experiments and achieve state-of-the-art results on two public datasets.

One of the main challenges in multi-source domain adaptation is how to reduce the domain discrepancy between each source domain and a target domain, and then evaluate the domain relevance to determine how much knowledge should be transferred from different source domains to the target domain. However, most prior approaches barely consider both discrepancies and relevance among domains. In this paper, we propose an algorithm, called Iterative Refinement based on Feature Selection and the Wasserstein distance (IRFSW), to solve semi-supervised domain adaptation with multiple sources. Specifically, IRFSW aims to explore both the discrepancies and relevance among domains in an iterative learning procedure, which gradually refines the learning performance until the algorithm stops. In each iteration, for each source domain and the target domain, we develop a sparse model to select features in which the domain discrepancy and training loss are reduced simultaneously. Then a classifier is constructed with the selected features of the source and labeled target data. After that, we exploit optimal transport over the selected features to calculate the transferred weights. The weight values are taken as the ensemble weights to combine the learned classifiers to control the amount of knowledge transferred from source domains to the target domain. Experimental results validate the effectiveness of the proposed method.

Scene flow in 3D point clouds plays an important role in understanding dynamic environments. Although significant advances have been made by deep neural networks, the performance is far from satisfactory as only per-point translational motion is considered, neglecting the constraints of the rigid motion in local regions. To address the issue, we propose to introduce the motion consistency to force the smoothness among neighboring points. In addition, constraints on the rigidity of the local transformation are also added by sharing unique rigid motion parameters for all points within each local region. To this end, a high-order CRFs based relation module (Con-HCRFs) is deployed to explore both point-wise smoothness and region-wise rigidity. To empower the CRFs to have a discriminative unary term, we also introduce a position-aware flow estimation module to be incorporated into the Con-HCRFs. Comprehensive experiments on FlyingThings3D and KITTI show that our proposed framework (HCRF-Flow) achieves state-of-the-art performance and significantly outperforms previous approaches substantially.

Due to the scarcity of annotated scene flow data, self-supervised scene flow learning in point clouds has attracted increasing attention. In the self-supervised manner, establishing correspondences between two point clouds to ap-proximate scene flow is an effective approach. Previous methods often obtain correspondences by applying point-wise matching that only takes the distance on 3D point co-ordinates into account, introducing two critical issues: (1) it overlooks other discriminative measures, such as color and surface normal, which often bring fruitful clues for ac-curate matching; and (2) it often generates sub-par performance, as the matching is operated in an unconstrained situation, where multiple points can be ended up with the same corresponding point. To address the issues, we formulate this matching task as an optimal transport problem. The output optimal assignment matrix can be utilized to guide the generation of pseudo ground truth. In this optimal transport, we design the transport cost by considering multiple descriptors and encourage one-to-one matching by mass equality constraints. Also, constructing a graph on the points, a random walk module is introduced to encourage the local consistency of the pseudo labels. Comprehensive experiments on FlyingThings3D and KITTI show that our method achieves state-of-the-art performance among self-supervised learning methods. Our self-supervised method even performs on par with some supervised learning approaches, although we do not need any ground truth flow for training.

Human multimodal emotion recognition involves time-series data of different modalities, such as natural language, visual motions, and acoustic behaviors. Due to the variable sampling rates for sequences from different modalities, the collected multimodal streams are usually unaligned. The asynchrony across modalities increases the difficulty on conducting efficient multimodal fusion. Hence, this work mainly focuses on multimodal fusion from unaligned multimodal sequences. To this end, we propose the Progressive Modality Reinforcement (PMR) approach based on the recent advances of crossmodal transformer. Our approach introduces a message hub to exchange information with each modality. The message hub sends common messages to each modality and reinforces their features via crossmodal attention. In turn, it also collects the reinforced features from each modality and uses them to generate a reinforced common message. By repeating the cycle process, the common message and the modalities’ features can progressively complement each other. Finally, the reinforced features are used to make predictions for human emotion. Comprehensive experiments on different human multimodal emotion recognition benchmarks clearly demonstrate the superiority of our approach.

Garment transfer shows great potential in realistic applications with the goal of transfering outfits across different people images. However, garment transfer between images with heavy misalignments or severe occlusions still remains as a challenge. In this work, we propose Complementary Transfering Network (CT-Net) to adaptively model different levels of geometric changes and transfer outfits between different people. In specific, CT-Net consists of three modules: i) A complementary warping module first estimates two complementary warpings to transfer the desired clothes in different granularities. ii) A layout prediction module is proposed to predict the target layout, which guides the preservation or generation of the body parts in the synthesized images. iii) A dynamic fusion module adaptively combines the advantages of the complementary warpings to render the garment transfer results. Extensive experiments conducted on DeepFashion dataset demonstrate that our network synthesizes high-quality garment transfer images and significantly outperforms the state-of-art methods both qualitatively and quantitatively. Our source code will be available online.

Few-shot class-incremental learning (FSCIL) aims to design machine learning algorithms that can continually learn new concepts from a few data points, without forgetting knowledge of old classes. The difficulty lies in that limited data from new classes not only lead to significant overfitting issues but also exacerbate the notorious catastrophic forgetting problems. Moreover, as training data come in sequence in FSCIL, the learned classifier can only provide discriminative information in individual sessions, while FSCIL requires all classes to be involved for evaluation. In this paper, we address the FSCIL problem from two aspects. First, we adopt a simple but effective decoupled learning strategy of representations and classifiers that only the classifiers are updated in each incremental session, which avoids knowledge forgetting in the representations. By doing so, we demonstrate that a pre-trained backbone plus a non-parametric class mean classifier can beat state-of-the-art methods. Second, to make the classifiers learned on individual sessions applicable to all classes, we propose a Continually Evolved Classifier (CEC) that employs a graph model to propagate context information between classifiers for adaptation. To enable the learning of CEC, we design a pseudo incremental learning paradigm that episodically constructs a pseudo incremental learning task to optimize the graph parameters by sampling data from the base dataset. Experiments on three popular benchmark datasets, including CIFAR100, miniImageNet, and Caltech-USCD Birds-200-2011 (CUB200), show that our method significantly outperforms the baselines and sets new state-of-the-art results with remarkable advantages.

Few-shot learning aims to adapt knowledge learned from previous tasks to novel tasks with only a limited amount of labeled data. Research literature on few-shot learning exhibits great diversity, while different algorithms often excel at different few-shot learning scenarios. It is therefore tricky to decide which learning strategies to use under different task conditions. Inspired by the recent success in Automated Machine Learning literature (AutoML), in this paper, we present Meta Navigator, a framework that attempts to solve the aforementioned limitation in few-shot learning by seeking a higher-level strategy and proffer to automate the selection from various few-shot learning designs. The goal of our work is to search for good parameter adaptation policies that are applied to different stages in the network for few-shot classification. We present a search space that covers many popular few-shot learning algorithms in the literature, and develop a differentiable searching and decoding algorithm based on meta-learning that supports gradient-based optimization. We demonstrate the effectiveness of our searching-based method on multiple benchmark datasets. Extensive experiments show that our approach significantly outperforms baselines and demonstrates performance advantages over many state-of-the-art methods.

Videos flow as the mixture of language, acoustic, and vision modalities. A thorough video understanding needs to fuse time-series data of different modalities for prediction. Due to the variable receiving frequency for sequences from each modality, there usually exists inherent asynchrony across the collected multimodal streams. Towards an efficient multimodal fusion from asynchronous multimodal streams, we need to model the correlations between elements from different modalities. The recent Multimodal Transformer (MulT) approach extends the self-attention mechanism of the original Transformer network to learn the crossmodal dependencies between elements. However, the direct replication of self-attention will suffer from the distribution mismatch across different modality features. As a result, the learnt crossmodal dependencies can be unreliable. Motivated by this observation, this work proposes the Modality-Invariant Crossmodal Attention (MICA) approach towards learning crossmodal interactions over modality-invariant space in which the distribution mismatch between different modalities is well bridged. To this end, both the marginal distribution and the elements with high-confidence correlations are aligned over the common space of the query and key vectors which are computed from different modalities. Experiments on three standard benchmarks of multimodal video understanding clearly validate the superiority of our approach.

Recently, self-supervised learning (SSL) has been proved very effective and it can help boost the performance in learning representations from unlabeled data in the image domain. Yet, very little is explored about its usefulness in 3D skeleton-based action recognition understanding. Directly applying existing SSL techniques for 3D skeleton learning, however, suffers from trivial solutions and imprecise representations. To tackle these drawbacks, we consider perceiving the consistency and continuity of motion at different playback speeds are two critical issues. To this end, we propose a novel SSL method to learn the 3D skeleton representation in an efficacious way. Specifically, by constructing a positive clip (speed-changed) and a negative clip (motion-broken) of the sampled action sequence, we encourage the positive pairs closer while pushing the negative pairs to force the network to learn the intrinsic dynamic motion consistency information. Moreover, to enhance the learning features, skeleton interpolation is further exploited to model the continuity of human skeleton data. To validate the effectiveness of the proposed method, extensive experiments are conducted on Kinetics, NTU60, NTU120, and PKUMMD datasets with several alternative network architectures. Experimental evaluations demonstrate the superiority of our approach and through which, we can gain significant performance improvement without using extra labeled data.

Data augmentation is vital for deep learning neural networks. By providing massive training samples, it helps to improve the generalization ability of the model. Weakly supervised semantic segmentation (WSSS) is a challenging problem that has been deeply studied in recent years, conventional data augmentation approaches for WSSS usually employ geometrical transformations, random cropping and color jittering. However, merely increasing the same contextual semantic data does not bring much gain to the networks to distinguish the objects, e.g., the correct image-level classification of "aeroplane" may be not only due to the recognition of the object itself, but also its co-occurrence context like "sky", which will cause the model to focus less on the object features. To this end, we present a Context Decoupling Augmentation (CDA) method, to change the inherent context in which the objects appear and thus drive the network to remove the dependence between object instances and contextual information. To validate the effectiveness of the proposed method, extensive experiments on PASCAL VOC 2012 and COCO datasets with several alternative network architectures demonstrate that CDA can boost various popular WSSS methods to the new state-of-the-art by a large margin. Code is available at https://github.com/suyukun666/CDA

Currently, the state-of-the-art methods treat few-shot semantic segmentation task as a conditional foreground-background segmentation problem, assuming each class is independent. In this paper, we introduce the concept of meta-class, which is the meta information (e.g. certain middle-level features) shareable among all classes. To explicitly learn meta-class representations in few-shot segmentation task, we propose a novel Meta-class Memory based few-shot segmentation method (MM-Net), where we introduce a set of learnable memory embeddings to memorize the meta-class information during the base class training and transfer to novel classes during the inference stage. Moreover, for the k-shot scenario, we propose a novel image quality measurement module to select images from the set of support images. A high-quality class prototype could be obtained with the weighted sum of support image features based on the quality measure. Experiments on both PASCAL-5iand COCO datasets show that our proposed method is able to achieve state-of-the-art results in both 1-shot and 5-shot settings. Particularly, our proposed MM-Net achieves 37.5% mIoU on the COCO dataset in 1-shot setting, which is 5.1% higher than the previous state-of-the-art.

Self-supervised learning (SSL) has been proved very effective in learning representations from unlabeled data in language and vision domains. Yet, very few instrumental self-supervised approaches exist for 3D skeleton action understanding, and directly applying the existing SSL methods from other domains for skeleton action learning may suffer from misalignment of representations and some limitations. In this paper, we consider that a good representation learning encoder can distinguish the underlying features of different actions, which can make the similar motions closer while pushing the dissimilar motions away. There exists, however, some uncertainties in the skeleton actions due to the inherent ambiguity of 3D skeleton pose in different viewpoints or the sampling algorithm in contrastive learning, thus, it is ill-posed to differentiate the action features in the deterministic embedding space. To address these issues, we rethink the distance between action features and propose to model each action representation into the probabilistic embedding space to alleviate the uncertainties upon encountering the ambiguous 3D skeleton inputs. To validate the effectiveness of the proposed method, extensive experiments are conducted on Kinetics, NTU60, NTU120, and PKUMMD datasets with several alternative network architectures. Experimental evaluations demonstrate the superiority of our approach and through which, we can gain significant performance improvement without using extra labeled data.

This paper investigates an open research task of text-to-image synthesis for automatically generating or manipulating images from text descriptions. Prevailing methods mainly take the textual descriptions as the conditional input for the GAN generation, and need to train different models for the text-guided image generation and manipulation tasks. In this paper, we propose a novel unified framework of Cycle-consistent Inverse GAN (CI-GAN) for both text-to-image generation and text-guided image manipulation tasks. Specifically, we first train a GAN model without text input, aiming to generate images with high diversity and quality. Then we learn a GAN inversion model to convert the images back to the GAN latent space and obtain the inverted latent codes for each image, where we introduce the cycle-consistency training to learn more robust and consistent inverted latent codes. We further uncover the semantics of the latent space of the trained GAN model, by learning a similarity model between text representations and the latent codes. In the text-guided optimization module, we can generate images with the desired semantic attributes through optimization on the inverted latent codes. Extensive experiments on the Recipe1M and CUB datasets validate the efficacy of our proposed framework.

With the development of Generative Adversarial Network, image-based virtual try-on methods have made great progress. However, limited work has explored the task of video-based virtual try-on while it is important in real-world applications. Most existing video-based virtual try-on methods usually require clothing templates and they can only generate blurred and low-resolution results. To address these challenges, we propose a Memory-based Video virtual Try-On Network (MV-TON), which seamlessly transfers desired clothes to a target person without using any clothing templates and generates high-resolution realistic videos. Specifically, MV-TON consists of two modules: 1) a try-on module that transfers the desired clothes from model images to frame images by pose alignment and region-wise replacing of pixels; 2) a memory refinement module that learns to embed the existing generated frames into the latent space as external memory for the following frame generation. Experimental results show the effectiveness of our method in the video virtual try-on task and its superiority over other existing methods.

3D pose transfer is one of the most challenging 3D generation tasks. It aims to transfer the pose of a source mesh to a target mesh and keep the identity (e.g., body shape) of the target mesh. Some previous works require key point annotations to build reliable correspondence between the source and target meshes, while other methods do not consider any shape correspondence between sources and targets, which leads to limited generation quality. In this work, we propose a correspondence-refinement network to achieve the 3D pose transfer for both human and animal meshes. The correspondence between source and target meshes is first established by solving an optimal transport problem. Then, we warp the source mesh according to the dense correspondence and obtain a coarse warped mesh. The warped mesh will be better refined with our proposed Elastic Instance Normalization, which is a conditional normalization layer and can help to generate high-quality meshes. Extensive experimental results show that the proposed architecture can effectively transfer the poses from source to target meshes and produce better results with satisfied visual performance than state-of-the-art methods.

Semi-supervised video object segmentation is a task of propagating instance masks given in the first frame to the entire video. It is a challenging task since it usually suffers from heavy occlusions, large deformation, and large variations of objects. To alleviate these problems, many existing works apply time-consuming techniques such as fine-tuning, post-processing, or extracting optical flow, which makes them intractable for online segmentation. In our work, we focus on online semi-supervised video object segmentation. We propose a GCSeg (Guided Co-Segmentation) Network which is mainly composed of a Reference Module and a Co-segmentation Module, to simultaneously incorporate the short-term, middle-term, and long-term temporal inter-frame relationships. Moreover, we propose an Adaptive Search Strategy to reduce the risk of propagating inaccurate segmentation results in subsequent frames. Our GCSeg network achieves state-of-the-art performance on online semi-supervised video object segmentation on Davis 2016 and Davis 2017 datasets.

Semantic segmentation, which aims to acquire pixel-level understanding about images, is among the key components in computer vision. To train a good segmentation model for real-world images, it usually requires a huge amount of time and labor effort to obtain sufficient pixel-level annotations of real-world images beforehand. To get rid of such a nontrivial burden, one can use simulators to automatically generate synthetic images that inherently contain full pixel-level annotations and use them to train a segmentation model for the real-world images. However, training with synthetic images usually cannot lead to good performance due to the domain difference between the synthetic images (i.e., source domain) and the real-world images (i.e., target domain). To deal with this issue, a number of unsupervised domain adaptation (UDA) approaches have been proposed, where no labeled real-world images are available. Different from those methods, in this work, we conduct a pioneer attempt by using easy-to-collect image-level annotations for target images to improve the performance of cross-domain segmentation. Specifically, we leverage those image-level annotations to construct curriculums for the domain adaptation problem. The curriculums describe multi-level properties of the target domain, including label distributions over full images, local regions and single pixels. Since image annotations are “weak” labels compared to pixel annotations for segmentation, we coin this new problem as weakly-supervised cross-domain segmentation. Comprehensive experiments on the GTA5→ Cityscapes and SYNTHIA→ Cityscapes settings demonstrate the effectiveness of our method over the existing state-of-the-art baselines.

The Internet of Things (IoT) will be a main data generation infrastructure for achieving better system intelligence. This article considers the design and implementation of a practical privacy-preserving collaborative learning scheme, in which a curious learning coordinator trains a better machine learning model based on the data samples contributed by a number of IoT objects, while the confidentiality of the raw forms of the training data is protected against the coordinator. Existing distributed machine learning and data encryption approaches incur significant computation and communication overhead, rendering them ill-suited for resource-constrained IoT objects. We study an approach that applies independent random projection at each IoT object to obfuscate data and trains a deep neural network at the coordinator based on the projected data from the IoT objects. This approach introduces light computation overhead to the IoT objects and moves most workload to the coordinator that can have sufficient computing resources. Although the independent projections performed by the IoT objects address the potential collusion between the curious coordinator and some compromised IoT objects, they significantly increase the complexity of the projected data. In this article, we leverage the superior learning capability of deep learning in capturing sophisticated patterns to maintain good learning performance. Extensive comparative evaluation shows that this approach outperforms other lightweight approaches that apply additive noisification for differential privacy and/or support vector machines for learning in the applications with light to moderate data pattern complexities.

We present a simple yet effective progressive self-guided loss function to facilitate deep learning-based salient object detection (SOD) in images. The saliency maps produced by the most relevant works still suffer from incomplete predictions due to the internal complexity of salient objects. Our proposed progressive self-guided loss simulates a morphological closing operation on the model predictions for progressively creating auxiliary training supervisions to step-wisely guide the training process. We demonstrate that this new loss function can guide the SOD model to highlight more complete salient objects step-by-step and meanwhile help to uncover the spatial dependencies of the salient object pixels in a region growing manner. Moreover, a new feature aggregation module is proposed to capture multi-scale features and aggregate them adaptively by a branch-wise attention mechanism. Benefiting from this module, our SOD framework takes advantage of adaptively aggregated multi-scale features to locate and detect salient objects effectively. Experimental results on several benchmark datasets show that our loss function not only advances the performance of existing SOD models without architecture modification but also helps our proposed framework to achieve state-of-the-art performance.

Image co-segmentation is an active computer vision task that aims to segment the common objects from a set of images. Recently, researchers design various learning-based algorithms to undertake the co-segmentation task. The main difficulty in this task is how to effectively transfer information between images to make conditional predictions. In this paper, we present CycleSegNet, a novel framework for the co-segmentation task. Our network design has two key components: a region correspondence module which is the basic operation for exchanging information between local image regions, and a cycle refinement module, which utilizes ConvLSTMs to progressively update image representations and exchange information in a cycle and iterative manner. Extensive experiments demonstrate that our proposed method significantly outperforms the state-of-the-art methods on four popular benchmark datasets - PASCAL VOC dataset, MSRC dataset, Internet dataset, and iCoseg dataset, by 2.6%, 7.7%, 2.2%, and 2.9%, respectively.

Over the past few years, state-of-the-art image segmentation algorithms are based on deep convolutional neural networks. To render a deep network with the ability to understand a concept, humans need to collect a large amount of pixel-level annotated data to train the models, which is time-consuming and tedious. Recently, few-shot segmentation is proposed to solve this problem. Few-shot segmentation aims to learn a segmentation model that can be generalized to novel classes with only a few training images. In this paper, we propose a cross-reference network (CRNet) for few-shot segmentation. Unlike previous works which only predict the mask in the query image, our proposed model concurrently makes predictions for both the support image and the query image. With a cross-reference mechanism, our network can better find the co-occurrent objects in two images, thus helping the few-shot segmentation task. We also develop a mask refinement module to recurrently refine the prediction of the foreground regions. For the k-shot learning, we propose to finetune parts of networks to take advantage of multiple labeled support images. Experiments on the PASCAL VOC 2012 dataset show that our network achieves state-of-the-art performance.

Exploiting photo-realistic synthetic data to train semantic segmentation models has received increasing attention over the past years. However, the domain mismatch between synthetic and real images will cause a significant performance drop when the model trained with synthetic images is directly applied to real-world scenarios. In this paper, we propose a new domain adaptation approach, called Pivot Interaction Transfer (PIT). Our method mainly focuses on constructing pivot information that is common knowledge shared across domains as a bridge to promote the adaptation of semantic segmentation model from synthetic domains to real-world domains. Specifically, we first infer the image-level category information about the target images, which is then utilized to facilitate pixel-level transfer for semantic segmentation, with the assumption that the interactive relation between the image-level category information and the pixel-level semantic information is invariant across domains. To this end, we propose a novel multi-level region expansion mechanism that aligns both the image-level and pixel-level information. Comprehensive experiments on the adaptation from both GTAV and SYNTHIA to Cityscapes clearly demonstrate the superiority of our method.

Point clouds are useful in many applications like autonomous driving and robotics as they provide natural 3D information of the surrounding environments. While there are extensive research on 3D point clouds, scene understanding on 4D point clouds, a series of consecutive 3D point clouds frames, is an emerging topic and yet under-investigated. With 4D point clouds (3D point cloud videos), robotic systems could enhance their robustness by leveraging the temporal information from previous frames. However, the existing semantic segmentation methods on 4D point clouds suffer from low precision due to the spatial and temporal information loss in their network structures. In this paper, we propose SpSequenceNet to address this problem. The network is designed based on 3D sparse convolution. And we introduce two novel modules, a cross-frame global attention module and a cross-frame local interpolation module, to capture spatial and temporal information in 4D point clouds. We conduct extensive experiments on SemanticKITTI, and achieve the state-of-the-art result of 43.1% on mIoU, which is 1.5% higher than the previous best approach.

Point clouds provide intrinsic geometric information and surface context for scene understanding. Existing methods for point cloud segmentation require a large amount of fully labeled data. Using advanced depth sensors, collection of large scale 3D dataset is no longer a cumbersome process. However, manually producing point-level label on the large scale dataset is time and labor-intensive. In this paper, we propose a weakly supervised approach to predict point-level results using weak labels on 3D point clouds. We introduce our multi-path region mining module to generate pseudo point-level labels from a classification network trained with weak labels. It mines the localization cues for each class from various aspects of the network feature using different attention modules. Then, we use the point-level pseudo label to train a point cloud segmentation network in a fully supervised manner. To the best of our knowledge, this is the first method that uses cloud-level weak labels on raw 3D space to train a point cloud semantic segmentation network. In our setting, the 3D weak labels only indicate the classes that appeared in our input sample. We discuss both scene- and subcloud-level weakly labels on raw 3D point cloud data and perform in-depth experiments on them. On ScanNet dataset, our result trained with subcloud-level labels is compatible with some fully supervised methods.

Fully supervised object detection has achieved great success in recent years. However, abundant bounding boxes annotations are needed for training a detector for novel classes. To reduce the human labeling effort, we propose a novel webly supervised object detection (WebSOD) method for novel classes which only requires the web images without further annotations. Our proposed method combines bottom-up and top-down cues for novel class detection. Within our approach, we introduce a bottom-up mechanism based on the well-trained fully supervised object detector (i.e. Faster RCNN) as an object region estimator for web images by recognizing the common objectiveness shared by base and novel classes. With the estimated regions on the web images, we then utilize the top-down attention cues as the guidance for region classification. Furthermore, we propose a residual feature refinement (RFR) block to tackle the domain mismatch between web domain and the target domain. We demonstrate our proposed method on PASCAL VOC dataset with three different novel/base splits. Without any target-domain novel-class images and annotations, our proposed webly supervised object detection model is able to achieve promising performance for novel classes. Moreover, we also conduct transfer learning experiments on large scale ILSVRC 2013 detection dataset and achieve state-of-the-art performance.

In this paper, we address the few-shot classification task from a new perspective of optimal matching between image regions. We adopt the Earth Mover's Distance (EMD) as a metric to compute a structural distance between dense image representations to determine image relevance. The EMD generates the optimal matching flows between structural elements that have the minimum matching cost, which is used to represent the image distance for classification. To generate the important weights of elements in the EMD formulation, we design a cross-reference mechanism, which can effectively minimize the impact caused by the cluttered background and large intra-class appearance variations. To handle k-shot classification, we propose to learn a structured fully connected layer that can directly classify dense image representations with the EMD. Based on the implicit function theorem, the EMD can be inserted as a layer into the network for end-to-end training. We conduct comprehensive experiments to validate our algorithm and we set new state-of-the-art performance on four popular few-shot classification benchmarks, namely miniImageNet, tieredImageNet, Fewshot-CIFAR100 (FC100) and Caltech-UCSD Birds-200-2011 (CUB).

In the weakly supervised segmentation task with only image-level labels, a common step in many existing algorithms is first to locate the image regions corresponding to each existing class with the Class Activation Maps (CAMs), and then generate the pseudo ground truth masks based on the CAMs to train a segmentation network in the fully supervised manner. The quality of the CAMs has a crucial impact on the performance of the segmentation model. We propose to improve the CAMs from a novel graph perspective. We model paired images containing common classes with a bipartite graph and use the maximum matching algorithm to locate corresponding areas in two images. The matching areas are then used to refine the predicted object regions in the CAMs. The experiments on Pascal VOC 2012 dataset show that our network can effectively boost the performance of the baseline model and achieves new state-of-the-art performance.

Recently, very deep convolutional neural networks (CNNs) have shown outstanding performance in object recognition and have also been the first choice for dense prediction problems such as semantic segmentation and depth estimation. However, repeated subsampling operations like pooling or convolution striding in deep CNNs lead to a significant decrease in the initial image resolution. Here, we present RefineNet, a generic multi-path refinement network that explicitly exploits all the information available along the down-sampling process to enable high-resolution prediction using long-range residual connections. In this way, the deeper layers that capture high-level semantic features can be directly refined using fine-grained features from earlier convolutions. The individual components of RefineNet employ residual connections following the identity mapping mindset, which allows for effective end-to-end training. Further, we introduce chained residual pooling, which captures rich background context in an efficient manner. We carry out comprehensive experiments on semantic segmentation which is a dense classification problem and achieve good performance on seven public datasets. We further apply our method for depth estimation and demonstrate the effectiveness of our method on dense regression problems.

In weakly supervised object detection, most existing approaches are proposed for images. Without box-level annotations, these methods cannot accurately locate objects. Considering an object may show different motion from its surrounding objects or background, we leverage motion information to improve the detection accuracy. However, the motion pattern of an object is complex. Different parts of an object may have different motion patterns, which poses challenges in exploring motion information for object localization. Directly using motion information may degrade the localization performance. To overcome these issues, we propose a Motion Context Network (MC-Net) in this letter. Our method generates motion context features by exploiting neighborhood motion correlation information on moving regions. These motion context features are then incorporated with image information to improve the detection accuracy. Furthermore, we propose a temporal aggregation module, which aggregates features across frames to enhance the feature representation at the current frame. Experiments are carried out on ImageNet VID, which shows that our MC-Net significantly improves the performance of the image based baseline method (37.4% mAP v.s. 29.8% mAP).

Depth is beneficial for salient object detection (SOD) for its additional saliency cues. Existing RGBD SOD methods focus on tailoring complicated cross-modal fusion topologies, which although achieve encouraging performance, are with a high risk of over-fitting and ambiguous in studying cross-modal complementarity. Different from these conventional approaches combining cross-modal features entirely without differentiating, we concentrate our attention on decoupling the diverse cross-modal complements to simplify the fusion process and enhance the fusion sufficiency. We argue that if cross-modal heterogeneous representations can be disentangled explicitly, the cross-modal fusion process can hold less uncertainty, while enjoying better adaptability. To this end, we design a disentangled cross-modal fusion network to expose structural and content representations from both modalities by cross-modal reconstruction. For different scenes, the disentangled representations allow the fusion module to easily identify, and incorporate desired complements for informative multi-modal fusion. Extensive experiments show the effectiveness of our designs and a large outperformance over state-of-the-art methods.

Object segmentation and object tracking are fundamental research areas in the computer vision community. These two topics are difficult to handle some common challenges, such as occlusion, deformation, motion blur, scale variation, and more. The former contains heterogeneous object, interacting object, edge ambiguity, and shape complexity; the latter suffers from difficulties in handling fast motion, out-of-view, and real-time processing. Combining the two problems of Video Object Segmentation and Tracking (VOST) can overcome their respective difficulties and improve their performance. VOST can be widely applied to many practical applications such as video summarization, high definition video compression, human computer interaction, and autonomous vehicles. This survey aims to provide a comprehensive review of the state-of-the-art VOST methods, classify these methods into different categories, and identify new trends. First, we broadly categorize VOST methods into Video Object Segmentation (VOS) and Segmentation-based Object Tracking (SOT). Each category is further classified into various types based on the segmentation and tracking mechanism. Moreover, we present some representative VOS and SOT methods of each time node. Second, we provide a detailed discussion and overview of the technical characteristics of the different methods. Third, we summarize the characteristics of the related video dataset and provide a variety of evaluation metrics. Finally, we point out a set of interesting future works and draw our own conclusions.

Recently, with the advent of deep convolutional neural networks (DCNN), the improvements in visual saliency prediction research are impressive. One possible direction to approach the next improvement is to fully characterize the multi-scale saliency-influential factors with a computationally-friendly module in DCNN architectures. In this work, we propose an end-to-end dilated inception network (DINet) for visual saliency prediction. It captures multi-scale contextual features effectively with very limited extra parameters. Instead of utilizing parallel standard convolutions with different kernel sizes as the existing inception module, our proposed dilated inception module (DIM) uses parallel dilated convolutions with different dilation rates which can significantly reduce the computation load while enriching the diversity of receptive fields in feature maps. Moreover, the performance of our saliency model is further improved by using a set of linear normalization-based probability distribution distance metrics as loss functions. As such, we can formulate saliency prediction as a global probability distribution prediction task for better saliency inference instead of a pixel-wise regression problem. Experimental results on several challenging saliency benchmark datasets demonstrate that our DINet with proposed loss functions can achieve state-of-the-art performance with shorter inference time.

It is challenging to detect curve texts due to their irregular shapes and varying sizes. In this paper, we first investigate the deficiency of the existing curve detection methods and then propose a novel Conditional Spatial Expansion (CSE) mechanism to improve the performance of curve text detection. Instead of regarding the curve text detection as a polygon regression or a segmentation problem, we treat it as a region expansion process. Our CSE starts with a seed arbitrarily initialized within a text region and progressively merges neighborhood regions based on the extracted local features by a CNN and contextual information of merged regions. The CSE is highly parameterized and can be seamlessly integrated into existing object detection frameworks. Enhanced by the data-dependent CSE mechanism, our curve text detection system provides robust instance-level text region extraction with minimal post-processing. The analysis experiment shows that our CSE can handle texts with various shapes, sizes, and orientations, and can effectively suppress the false-positives coming from text-like textures or unexpected texts included in the same RoI. Compared with the existing curve text detection algorithms, our method is more robust and enjoys a simpler processing flow. It also creates a new state-of-art performance on curve text benchmarks with Fscore of up to 78.4%.

Recent progress in semantic segmentation is driven by deep Convolutional Neural Networks and large-scale labeled image datasets. However, data labeling for pixel-wise segmentation is tedious and costly. Moreover, a trained model can only make predictions within a set of pre-defined classes. In this paper, we present CANet, a class-agnostic segmentation network that performs few-shot segmentation on new classes with only a few annotated images available. Our network consists of a two-branch dense comparison module which performs multi-level feature comparison between the support image and the query image, and an iterative optimization module which iteratively refines the predicted results. Furthermore, we introduce an attention mechanism to effectively fuse information from multiple support examples under the setting of k-shot learning. Experiments on PASCAL VOC 2012 show that our method achieves a mean Intersection-over-Union score of 55.4% for 1-shot segmentation and 57.1% for 5-shot segmentation, outperforming state-of-the-art methods by a large margin of 14.6% and 13.2%, respectively.

Embodied Question Answering (EQA) is a newly defined research area where an agent is required to answer the user's questions by exploring the real world environment. It has attracted increasing research interests due to its broad applications in automatic driving system, in-home robots, and personal assistants. Most of the existing methods perform poorly in terms of answering and navigation accuracy due to the absence of local details and vulnerability to the ambiguity caused by complicated vision conditions. To tackle these problems, we propose a segmentation based visual attention mechanism for Embodied Question Answering. Firstly, We extract the local semantic features by introducing a novel high-speed video segmentation framework. Then by the guide of extracted semantic features, a bottom-up visual attention mechanism is proposed for the Visual Question Answering (VQA) sub-task. Further, a feature fusion strategy is proposed to guide the training of the navigator without much additional computational cost. The ablation experiments show that our method boosts the performance of VQA module by 4.2% (68.99% vs 64.73%) and leads to 3.6% (48.59% vs 44.98%) overall improvement in EQA accuracy.

One-shot image segmentation aims to undertake the segmentation task of a novel class with only one training image available. The difficulty lies in that image segmentation has structured data representations, which yields a many-to-many message passing problem. Previous methods often simplify it to a one-to-many problem by squeezing support data to a global descriptor. However, a mixed global representation drops the data structure and information of individual elements. In this paper, we propose to model structured segmentation data with graphs and apply attentive graph reasoning to propagate label information from support data to query data. The graph attention mechanism could establish the element-to-element correspondence across structured data by learning attention weights between connected graph nodes. To capture correspondence at different semantic levels, we further propose a pyramid-like structure that models different sizes of image regions as graph nodes and undertakes graph reasoning at different levels. Experiments on PASCAL VOC 2012 dataset demonstrate that our proposed network significantly outperforms the baseline method and leads to new state-of-the-art performance on 1-shot and 5-shot segmentation benchmarks.

Most existing virtual try-on applications require clean clothes images. Instead, we present a novel virtual Try-On network, M2E-Try On Net, which transfers the clothes from a model image to a person image without the need of any clean product images. To obtain a realistic image of person wearing the desired model clothes, we aim to solve the following challenges: 1) non-rigid nature of clothes - we need to align poses between the model and the user; 2) richness in textures of fashion items - preserving the fine details and characteristics of the clothes is critical for photo-realistic transfer; 3) variation of identity appearances - it is required to fit the desired model clothes to the person identity seamlessly. To tackle these challenges, we introduce three key components, including the pose alignment network (PAN), the texture refinement network (TRN) and the fitting network (FTN). Since it is unlikely to gather image pairs of input person image and desired output image (i.e. person wearing the desired clothes), our framework is trained in a self-supervised manner to gradually transfer the poses and textures of the model's clothes to the desired appearance. In the experiments, we verify on the Deep Fashion dataset and MVC dataset that our method can generate photo-realistic images for the person to try-on the model clothes. Furthermore, we explore the model capability for different fashion items, including both upper and lower garments.

In this paper we focus on the problem of unsupervised domain adaptation for semantic segmentation. The previous works usually focus on adversarial learning either in pixel-level or feature-level. However, global structure knowledge is often neglected in the adversarial learning due to the possible reasons: First, the result of pixel-level adversarial learning does not necessarily preserve the semantic consistency of the input image. Second, global structure knowledge is not embedded to regularize the feature-level adversarial learning. In this work, we propose a framework for unsupervised domain adaptation in semantic segmentation which effectively incorporates pixel- level, feature-level adversarial learning and self-training strategy. Our framework embeds the global structure knowledge into the adversarial training step to tackle the problem of structure misalignment. Consequently, our proposed framework achieves the state-of-the-art semantic segmentation domain adaptation results on the task of transferring GTA5 to Cityscapes.

In this work we tackle the problem of unsupervised domain adaptation for semantic segmentation. One pipeline is to sequentially train image-translation model and the final task segmentation model. In such pipeline, image translation is aimed to generate the translated source-domain images which are visually similar to the target-domain images and then the final task model is trained using the translated images and its corresponding groundtruth. However, the visually optimal translated-images are not necessarily optimal for the final task of segmenting the target-domain images. Thus we propose a Task-in-all pipeline for unsupervised domain adaptation on semantic segmentation, which incorporates image translation and final segmentation task into an end-to-end training pipeline. Our aim is to generate the translated images which better assists the final task, instead of just being visually similar to the target domain images. We show that in the task of adapting from GTA5 to Cityscapes dataset, the segmentation performance of our Task-in-all pipeline outperforms the sequentially training pipeline, with simpler model structure and less training complexity.

Action recognition is an important yet challenging problem. We here present a novel method, multistage local fusion networks with residual connections, to boost the performance of video action recognition . In realistic videos, an action instance may have a long time span and some frames may suffer from deteriorated object appearance due to motion blur or video defocus. Our method enhances the per-frame representation by capturing information from neighboring frames. We propose a local fusion block which considers neighboring frames to capture appearance and local motion information for generating per-frame representation. Our local fusion is performed in a multistage manner allowing feature fusion from varying neighborhood sizes in the temporal dimension. We employ residual connections in the fusion blocks to enable effective gradient propagation through the whole network allowing effective end-to-end training. We achieve competitive results on two challenging and public available datasets, namely HMDB51 and UCF101, which shows the effectiveness of the proposed method.

In this paper, we propose a semantics-aware visual object tracking method, which introduces semantics into the tracking procedure and extends the model of an object with explicit semantics prior to enhancing the robustness of three key aspects of the tracking framework, i.e., appearance model, search scheme, and scale adaptation. We first present a semantic object proposal generation method for video sequences to generate high-quality category-oriented object proposals. Then, a hybrid semantics-aware tracking algorithm with semantic compatibility is proposed. This algorithm takes full advantages of globally sparse semantic object proposal prediction and locally dense prediction with a template model and semantic distractor-aware color appearance model. Furthermore, we propose to exploit semantics to localize object accurately via an energy minimization framework-based scale adaptation method, which jointly integrates dense location prior, instance-specific color, and category-specific semantic information. Extensive experiments are conducted on two widely used benchmarks, and the results demonstrate that our method achieves the state-of-the-art performance.

Weakly supervised semantic segmentation receives much research attention since it alleviates the need to obtain a large amount of dense pixel-wise ground-truth annotations for the training images. Compared with other forms of weak supervision, image labels are quite efficient to obtain. In this paper, we focus on the weakly supervised semantic segmentation with image label annotations. Recent progress for this task has been largely dependent on the quality of generated pseudo-annotations. In this paper, inspired by spatial neural-attention for image captioning, we propose a decoupled spatial neural attention network for generating pseudo-annotations. Our decoupled attention structure could simultaneously identify the object regions and localize the discriminative parts, which generates high-quality pseudo-annotations in one forward path. The generated pseudo-annotations lead to the segmentation results that achieve the state of the art in weakly supervised semantic segmentation.

A novel framework named Markov Clustering Network (MCN) is proposed for fast and robust scene text detection. MCN predicts instance-level bounding boxes by firstly converting an image into a Stochastic Flow Graph (SFG) and then performing Markov Clustering on this graph. Our method can detect text objects with arbitrary size and orientation without prior knowledge of object size. The stochastic flow graph encode objects' local correlation and semantic information. An object is modeled as strongly connected nodes, which allows flexible bottom-up detection for scale-varying and rotated objects. MCN generates bounding boxes without using Non-Maximum Suppression, and it can be fully parallelized on GPUs. The evaluation on public benchmarks shows that our method outperforms the existing methods by a large margin in detecting multioriented text objects. MCN achieves new state-of-art performance on challenging MSRA-TD500 dataset with precision of 0.88, recall of 0.79 and F-score of 0.83. Also, MCN achieves realtime inference with frame rate of 34 FPS, which is 1.5× speedup when compared with the fastest scene text detection algorithm.

Fully supervised methods for semantic segmentation require pixel-level class masks to train, the creation of which is expensive in terms of manual labour and time. In this work, we focus on weak supervision, developing a method for training a high-quality pixel-level classifier for semantic segmentation, using only image-level class labels as the provided ground-truth. Our method is formulated as a two-stage approach in which we first aim to create accurate pixel-level masks for the training images via a bootstrapping process, and then use these now-accurately segmented images as a proxy ground-truth in a more standard supervised setting. The key driver for our work is that in the target dataset we typically have reliable ground-truth image-level labels, while data crawled from the web may have unreliable labels, but can be filtered to comprise only easy images to segment, therefore having reliable boundaries. These two forms of information are complementary and we use this observation to build a novel bi-directional transfer learning framework. This framework transfers knowledge between two domains, target domain and web domain, bootstrapping the performance of weakly supervised semantic segmentation. Conducting experiments on the popular benchmark dataset PASCAL VOC 2012 based on both a VGG16 network and on ResNet50, we reach state-of-the-art performance with scores of 60.2% IoU and 63.9% IoU respectively1.

In this paper, we propose a novel MoNet model to deeply exploit motion cues for boosting video object segmentation performance from two aspects, i.e., frame representation learning and segmentation refinement. Concretely, MoNet exploits computed motion cue (i.e., optical flow) to reinforce the representation of the target frame by aligning and integrating representations from its neighbors. The new representation provides valuable temporal contexts for segmentation and improves robustness to various common contaminating factors, e.g., motion blur, appearance variation and deformation of video objects. Moreover, MoNet exploits motion inconsistency and transforms such motion cue into foreground/background prior to eliminate distraction from confusing instances and noisy regions. By introducing a distance transform layer, MoNet can effectively separate motion-inconstant instances/regions and thoroughly refine segmentation results. Integrating the proposed two motion exploitation components with a standard segmentation network, MoNet provides new state-of-the-art performance on three competitive benchmark datasets.

We present our approach for robotic perception in cluttered scenes that led to winning the recent Amazon Robotics Challenge (ARC) 2017. Next to small objects with shiny and transparent surfaces, the biggest challenge of the 2017 competition was the introduction of unseen categories. In contrast to traditional approaches which require large collections of annotated data and many hours of training, the task here was to obtain a robust perception pipeline with only few minutes of data acquisition and training time. To that end, we present two strategies that we explored. One is a deep metric learning approach that works in three separate steps: semantic-agnostic boundary detection, patch classification and pixel-wise voting. The other is a fully-supervised semantic segmentation approach with efficient dataset collection. We conduct an extensive analysis of the two methods on our ARC 2017 dataset. Interestingly, only few examples of each class are sufficient to fine-tune even very deep convolutional neural networks for this specific task.

We propose an approach for exploiting contextual information in semantic image segmentation, and particularly investigate the use of patch-patch context and patch-background context in deep CNNs. We formulate deep structured models by combining CNNs and Conditional Random Fields (CRFs) for learning the patch-patch context between image regions. Specifically, we formulate CNN-based pairwise potential functions to capture semantic correlations between neighboring patches. Efficient piecewise training of the proposed deep structured model is then applied in order to avoid repeated expensive CRF inference during the course of back propagation. For capturing the patch-background context, we show that a network design with traditional multi-scale image inputs and sliding pyramid pooling is very effective for improving performance. We perform comprehensive evaluation of the proposed method. We achieve new state-of-the-art performance on a number of challenging semantic segmentation datasets.

Crowd counting is an important task in computer vision, which has many applications in video surveillance. Although the regression-based framework has achieved great improvements for crowd counting, how to improve the discriminative power of image representation is still an open problem. Conventional holistic features used in crowd counting often fail to capture semantic attributes and spatial cues of the image. In this paper, we propose integrating semantic information into learning locality-aware feature (LAF) sets for accurate crowd counting. First, with the help of a convolutional neural network, the original pixel space is mapped onto a dense attribute feature map, where each dimension of the pixelwise feature indicates the probabilistic strength of a certain semantic class. Then, LAF built on the idea of spatial pyramids on neighboring patches is proposed to explore more spatial context and local information. Finally, the traditional vector of locally aggregated descriptor (VLAD) encoding method is extended to a more generalized form weighted-VLAD (W-VLAD) in which diverse coefficient weights are taken into consideration. Experimental results validate the effectiveness of our presented method.

We propose a new approach to image segmentation, which exploits the advantages of both conditional random fields (CRFs) and decision trees. In the literature, the potential functions of CRFs are mostly defined as a linear combination of some predefined parametric models, and then, methods, such as structured support vector machines, are applied to learn those linear coefficients. We instead formulate the unary and pairwise potentials as nonparametric forests-ensembles of decision trees, and learn the ensemble parameters and the trees in a unified optimization problem within the large-margin framework. In this fashion, we easily achieve nonlinear learning of potential functions on both unary and pairwise terms in CRFs. Moreover, we learn classwise decision trees for each object that appears in the image. Experimental results on several public segmentation data sets demonstrate the power of the learned nonlinear nonparametric potentials.

Recognizing the identities of people in everyday photos is still a very challenging problem for machine vision, due to issues such as non-frontal faces, changes in clothing, location, lighting. Recent studies have shown that rich relational information between people in the same photo can help in recognizing their identities. In this work, we propose to model the relational information between people as a sequence prediction task. At the core of our work is a novel recurrent network architecture, in which relational information between instances labels and appearance are modeled jointly. In addition to relational cues, scene context is incorporated in our sequence prediction model with no additional cost. In this sense, our approach is a unified framework for modeling both contextual cues and visual appearance of person instances. Our model is trained end-to-end with a sequence of annotated instances in a photo as inputs, and a sequence of corresponding labels as targets. We demonstrate that this simple but elegant formulation achieves state-of-the-art performance on the newly released People In Photo Albums (PIPA) dataset.

Recently, very deep convolutional neural networks (CNNs) have shown outstanding performance in object recognition and have also been the first choice for dense classification problems such as semantic segmentation. However, repeated subsampling operations like pooling or convolution striding in deep CNNs lead to a significant decrease in the initial image resolution. Here, we present RefineNet, a generic multi-path refinement network that explicitly exploits all the information available along the down-sampling process to enable high-resolution prediction using long-range residual connections. In this way, the deeper layers that capture high-level semantic features can be directly refined using fine-grained features from earlier convolutions. The individual components of RefineNet employ residual connections following the identity mapping mindset, which allows for effective end-to-end training. Further, we introduce chained residual pooling, which captures rich background context in an efficient manner. We carry out comprehensive experiments and set new state-of-the-art results on seven public datasets. In particular, we achieve an intersection-over-union score of 83.4 on the challenging PASCAL VOC 2012 dataset, which is the best reported result to date.

Semantic image segmentation is a fundamental task in image understanding. Per-pixel semantic labelling of an image benefits greatly from the ability to consider region consistency both locally and globally. However, many Fully Convolutional Network based methods do not impose such consistency, which may give rise to noisy and implausible predictions. We address this issue by proposing a dense multi-label network module that is able to encourage the region consistency at different levels. This simple but effective module can be easily integrated into any semantic segmentation systems. With comprehensive experiments, we show that the dense multi-label can successfully remove the implausible labels and clear the confusion so as to boost the performance of semantic segmentation systems.

Recent works on deep conditional random fields (CRFs) have set new records on many vision tasks involving structured predictions. Here, we propose a fully connected deep continuous CRF model with task-specific losses for both discrete and continuous labeling problems. We exemplify the usefulness of the proposed model on multi-class semantic labeling (discrete) and the robust depth estimation (continuous) problems. In our framework, we model both the unary and the pairwise potential functions as deep convolutional neural networks (CNNs), which are jointly learned in an end-to-end fashion. The proposed method possesses the main advantage of continuously valued CRFs, which is a closed-form solution for the maximum a posteriori (MAP) inference. To better take into account the quality of the predicted estimates during the cause of learning, instead of using the commonly employed maximum likelihood CRF parameter learning protocol, we propose task-specific loss functions for learning the CRF parameters. It enables direct optimization of the quality of the MAP estimates during the learning process. Specifically, we optimize the multi-class classification loss for the semantic labeling task and the Tukey's biweight loss for the robust depth estimation problem. Experimental results on the semantic labeling and robust depth estimation tasks demonstrate that the proposed method compare favorably against both baseline and state-of-the-art methods. In particular, we show that although the proposed deep CRF model is continuously valued, with the equipment of task-specific loss, it achieves impressive results even on discrete labeling tasks.

Recent advances in semantic image segmentation have mostly been achieved by training deep convolutional neural networks (CNNs). We show how to improve semantic segmentation through the use of contextual information, specifically, we explore 'patch-patch' context between image regions, and 'patch-background' context. For learning from the patch-patch context, we formulate Conditional Random Fields (CRFs) with CNN-based pairwise potential functions to capture semantic correlations between neighboring patches. Efficient piecewise training of the proposed deep structured model is then applied to avoid repeated expensive CRF inference for back propagation. For capturing the patch-background context, we show that a network design with traditional multi-scale image input and sliding pyramid pooling is effective for improving performance. Our experimental results set new state-of-the-art performance on a number of popular semantic segmentation datasets, including NYUDv2, PASCAL VOC 2012, PASCAL-Context, and SIFT-flow. In particular, we achieve an intersection-overunion score of 78:0 on the challenging PASCAL VOC 2012 dataset.

In this paper, we aim to learn a mapping (or embedding) from images to a compact binary space in which Hamming distances correspond to a ranking measure for the image retrieval task. We make use of a triplet loss because this has been shown to be most effective for ranking problems. However, training in previous works can be prohibitively expensive due to the fact that optimization is directly performed on the triplet space, where the number of possible triplets for training is cubic in the number of training examples. To address this issue, we propose to formulate high-order binary codes learning as a multi-label classification problem by explicitly separating learning into two interleaved stages. To solve the first stage, we design a large-scale high-order binary codes inference algorithm to reduce the high-order objective to a standard binary quadratic problem such that graph cuts can be used to efficiently infer the binary codes which serve as the labels of each training datum. In the second stage we propose to map the original image to compact binary codes via carefully designed deep convolutional neural networks (CNNs) and the hashing function fitting can be solved by training binary CNN classifiers. An incremental/interleaved optimization strategy is proffered to ensure that these two steps are interactive with each other during training for better accuracy. We conduct experiments on several benchmark datasets, which demonstrate both improved training time (by as much as two orders of magnitude) as well as producing state-of-the-art hashing for various retrieval tasks.

In this article, we tackle the problem of depth estimation from single monocular images. Compared with depth estimation using multiple images such as stereo depth perception, depth from monocular images is much more challenging. Prior work typically focuses on exploiting geometric priors or additional sources of information, most using hand-crafted features. Recently, there is mounting evidence that features from deep convolutional neural networks (CNN) set new records for various vision applications. On the other hand, considering the continuous characteristic of the depth values, depth estimation can be naturally formulated as a continuous conditional random field (CRF) learning problem. Therefore, here we present a deep convolutional neural field model for estimating depths from single monocular images, aiming to jointly explore the capacity of deep CNN and continuous CRF. In particular, we propose a deep structured learning scheme which learns the unary and pairwise potentials of continuous CRF in a unified deep CNN framework. We then further propose an equally effective model based on fully convolutional networks and a novel superpixel pooling method, which is about 10 times faster, to speedup the patch-wise convolutions in the deep model. With this more efficient model, we are able to design deeper networks to pursue better performance. Our proposed method can be used for depth estimation of general scenes with no geometric priors nor any extra information injected. In our case, the integral of the partition function can be calculated in a closed form such that we can exactly solve the log-likelihood maximization. Moreover, solving the inference problem for predicting depths of a test image is highly efficient as closed-form solutions exist. Experiments on both indoor and outdoor scene datasets demonstrate that the proposed method outperforms state-of-the-art depth estimation approaches.

We consider the problem of depth estimation from a single monocular image in this work. It is a challenging task as no reliable depth cues are available, e.g., stereo correspondences, motions etc. Previous efforts have been focusing on exploiting geometric priors or additional sources of information, with all using hand-crafted features. Recently, there is mounting evidence that features from deep convolutional neural networks (CNN) are setting new records for various vision applications. On the other hand, considering the continuous characteristic of the depth values, depth estimations can be naturally formulated into a continuous conditional random field (CRF) learning problem. Therefore, we in this paper present a deep convolutional neural field model for estimating depths from a single image, aiming to jointly explore the capacity of deep CNN and continuous CRF. Specifically, we propose a deep structured learning scheme which learns the unary and pairwise potentials of continuous CRF in a unified deep CNN framework. The proposed method can be used for depth estimations of general scenes with no geometric priors nor any extra information injected. In our case, the integral of the partition function can be analytically calculated, thus we can exactly solve the log-likelihood optimization. Moreover, solving the MAP problem for predicting depths of a new image is highly efficient as closed-form solutions exist. We experimentally demonstrate that the proposed method outperforms state-of-the-art depth estimation methods on both indoor and outdoor scene datasets.

Deep structured output learning shows great promise in tasks like semantic image segmentation. We proffer a new, efficient deep structured model learning scheme, in which we show how deep Convolutional Neural Networks (CNNs) can be used to directly estimate the messages in message passing inference for structured prediction with Conditional Random Fields CRFs). With such CNN message estimators, we obviate the need to learn or evaluate potential functions for message calculation. This confers significant efficiency for learning, since otherwise when performing structured learning for a CRF with CNN potentials it is necessary to undertake expensive inference for every stochastic gradient iteration. The network output dimension of message estimators is the same as the number of classes, rather than exponentially growing in the order of the potentials. Hence it is more scalable for cases that a large number of classes are involved. We apply our method to semantic image segmentation and achieve impressive performance, which demonstrates the effectiveness and usefulness of our CNN message learning method.

To build large-scale query-by-example image retrieval systems, embedding image features into a binary Hamming space provides great benefits. Supervised hashing aims to map the original features to compact binary codes that are able to preserve label based similarity in the binary Hamming space. Most existing approaches apply a single form of hash function, and an optimization process which is typically deeply coupled to this specific form. This tight coupling restricts the flexibility of those methods, and can result in complex optimization problems that are difficult to solve. In this work we proffer a flexible yet simple framework that is able to accommodate different types of loss functions and hash functions. The proposed framework allows a number of existing approaches to hashing to be placed in context, and simplifies the development of new problem-specific hashing methods. Our framework decomposes the hashing learning problem into two steps: binary code (hash bit) learning and hash function learning. The first step can typically be formulated as binary quadratic problems, and the second step can be accomplished by training a standard binary classifier. For solving large-scale binary code inference, we show how it is possible to ensure that the binary quadratic problems are submodular such that efficient graph cut methods may be used. To achieve efficiency as well as efficacy on large-scale high-dimensional data, we propose to use boosted decision trees as the hash functions, which are nonlinear, highly descriptive, and are very fast to train and evaluate. Experiments demonstrate that the proposed method significantly outperforms most state-of-the-art methods, especially on high-dimensional data.

Supervised hashing aims to map the original features to compact binary codes that are able to preserve label based similarity in the Hamming space. Non-linear hash functions have demonstrated their advantage over linear ones due to their powerful generalization capability. In the literature, kernel functions are typically used to achieve non-linearity in hashing, which achieve encouraging retrieval perfor- mance at the price of slow evaluation and training time. Here we propose to use boosted decision trees for achieving non-linearity in hashing, which are fast to train and evaluate, hence more suitable for hashing with high dimensional data. In our approach, we first propose sub-modular formulations for the hashing binary code inference problem and an efficient GraphCut based block search method for solving large-scale inference. Then we learn hash func- tions by training boosted decision trees to fit the binary codes. Experiments demonstrate that our proposed method significantly outperforms most state-of-the-art methods in retrieval precision and training time. Especially for high- dimensional data, our method is orders of magnitude faster than many methods in terms of training time.

Boosting is a method for learning a single accurate predictor by linearly combining a set of less accurate weak learners. Recently, structured learning has found many applications in computer vision. Inspired by structured support vector machines (SSVM), here we propose a new boosting algorithm for structured output prediction, which we refer to as StructBoost. StructBoost supports nonlinear structured learning by combining a set of weak structured learners. As SSVM generalizes SVM, our StructBoost generalizes standard boosting approaches such as AdaBoost, or LPBoost to structured learning. The resulting optimization problem of StructBoost is more challenging than SSVM in the sense that it may involve exponentially many variables and constraints. In contrast, for SSVM one usually has an exponential number of constraints and a cutting-plane method is used. In order to efficiently solve StructBoost, we formulate an equivalent 1-slack formulation and solve it using a combination of cutting planes and column generation. We show the versatility and usefulness of StructBoost on a range of problems such as optimizing the tree loss for hierarchical multi-class classification, optimizing the Pascal overlap criterion for robust visual tracking and learning conditional random field parameters for image segmentation.

Most existing approaches to hashing apply a single form of hash function, and an optimization process which is typically deeply coupled to this specific form. This tight coupling restricts the flexibility of the method to respond to the data, and can result in complex optimization problems that are difficult to solve. Here we propose a flexible yet simple framework that is able to accommodate different types of loss functions and hash functions. This framework allows a number of existing approaches to hashing to be placed in context, and simplifies the development of new problem-specific hashing methods. Our framework decomposes the hashing learning problem into two steps: hash bit learning and hash function learning based on the learned bits. The first step can typically be formulated as binary quadratic problems, and the second step can be accomplished by training standard binary classifiers. Both problems have been extensively studied in the literature. Our extensive experiments demonstrate that the proposed framework is effective, flexible and outperforms the state-of-the-art.

We propose efficient training methods (SBoost) for totally-corrective boosting based structured learning. The optimization of boosting method for structured learning is more challenging than the structured support vector machine. Basically, we propose smooth and convex formulation for boosting based structured learning, and develop approximate constraint generation together with column generation to solve the optimization with large number of constraints and variables. Because of the convexity and smoothness, the optimization in each generation iteration can be solved efficiently. We demonstrate some structured learning applications in computer vision using SBoost, including invariance learning for digit recognition, object detection and hierarchical image classification.

Fast nearest neighbor searching is becoming an increasingly important tool in solving many large-scale problems. Recently a number of approaches to learning data-dependent hash functions have been developed. In this work, we propose a column generation based method for learning data-dependent hash functions on the basis of proximity comparison information. Given a set of triplets that encode the pairwise proximity comparison information, our method learns hash functions that preserve the relative comparison relationships in the data as well as possible within the large-margin learning framework. The learning procedure is implemented using column generation and hence is named CGHash. At each iteration of the column generation procedure, the best hash function is selected. Unlike most other hashing methods, our method generalizes to new data points naturally; and has a training objective which is convex, thus ensuring that the global optimum can be identified. Experiments demonstrate that the proposed method learns compact binary codes and that its retrieval performance compares favorably with state-of-the-art methods when tested on a few benchmark datasets.