Hongxun Yao

Haosen Yang, Deng Huang, Bin Wen et al.

Masked autoencoders (MAEs) have emerged recently as art self-supervised spatiotemporal representation learners. Inheriting from the image counterparts, however, existing video MAEs still focus largely on static appearance learning whilst are limited in learning dynamic temporal information hence less effective for video downstream tasks. To resolve this drawback, in this work we present a motion-aware variant -- MotionMAE. Apart from learning to reconstruct individual masked patches of video frames, our model is designed to additionally predict the corresponding motion structure information over time. This motion information is available at the temporal difference of nearby frames. As a result, our model can extract effectively both static appearance and dynamic motion spontaneously, leading to superior spatiotemporal representation learning capability. Extensive experiments show that our MotionMAE outperforms significantly both supervised learning baseline and state-of-the-art MAE alternatives, under both domain-specific and domain-generic pretraining-then-finetuning settings. In particular, when using ViT-B as the backbone our MotionMAE surpasses the prior art model by a margin of 1.2% on Something-Something V2 and 3.2% on UCF101 in domain-specific pretraining setting. Encouragingly, it also surpasses the competing MAEs by a large margin of over 3% on the challenging video object segmentation task. The code is available at https://github.com/happy-hsy/MotionMAE.

Shuo Yang, Zhaopan Xu, Kai Wang et al.

As one of the most fundamental techniques in multimodal learning, cross-modal matching aims to project various sensory modalities into a shared feature space. To achieve this, massive and correctly aligned data pairs are required for model training. However, unlike unimodal datasets, multimodal datasets are extremely harder to collect and annotate precisely. As an alternative, the co-occurred data pairs (e.g., image-text pairs) collected from the Internet have been widely exploited in the area. Unfortunately, the cheaply collected dataset unavoidably contains many mismatched data pairs, which have been proven to be harmful to the model's performance. To address this, we propose a general framework called BiCro (Bidirectional Cross-modal similarity consistency), which can be easily integrated into existing cross-modal matching models and improve their robustness against noisy data. Specifically, BiCro aims to estimate soft labels for noisy data pairs to reflect their true correspondence degree. The basic idea of BiCro is motivated by that -- taking image-text matching as an example -- similar images should have similar textual descriptions and vice versa. Then the consistency of these two similarities can be recast as the estimated soft labels to train the matching model. The experiments on three popular cross-modal matching datasets demonstrate that our method significantly improves the noise-robustness of various matching models, and surpass the state-of-the-art by a clear margin.

Huicong Zhang, Haozhe Xie, Hongxun Yao

The key success factor of the video deblurring methods is to compensate for the blurry pixels of the mid-frame with the sharp pixels of the adjacent video frames. Therefore, mainstream methods align the adjacent frames based on the estimated optical flows and fuse the alignment frames for restoration. However, these methods sometimes generate unsatisfactory results because they rarely consider the blur levels of pixels, which may introduce blurry pixels from video frames. Actually, not all the pixels in the video frames are sharp and beneficial for deblurring. To address this problem, we propose the spatio-temporal deformable attention network (STDANet) for video delurring, which extracts the information of sharp pixels by considering the pixel-wise blur levels of the video frames. Specifically, STDANet is an encoder-decoder network combined with the motion estimator and spatio-temporal deformable attention (STDA) module, where motion estimator predicts coarse optical flows that are used as base offsets to find the corresponding sharp pixels in STDA module. Experimental results indicate that the proposed STDANet performs favorably against state-of-the-art methods on the GoPro, DVD, and BSD datasets.

Wei Yu, Kuiyuan Yang, Yalong Bai et al.

The gap between low-level visual signals and high-level semantics has been progressively bridged by continuous development of deep neural network (DNN). With recent progress of DNN, almost all image classification tasks have achieved new records of accuracy. To extend the ability of DNN to image retrieval tasks, we proposed a unified DNN model for image-query similarity calculation by simultaneously modeling image and query in one network. The unified DNN is named the cross space mapping (CSM) model, which contains two parts, a convolutional part and a query-embedding part. The image and query are mapped to a common vector space via these two parts respectively, and image-query similarity is naturally defined as an inner product of their mappings in the space. To ensure good generalization ability of the DNN, we learn weights of the DNN from a large number of click-through logs which consists of 23 million clicked image-query pairs between 1 million images and 11.7 million queries. Both the qualitative results and quantitative results on an image retrieval evaluation task with 1000 queries demonstrate the superiority of the proposed method.

Chonghuinan Wang, Zihan Chen, Yuxiang Wei et al.

Instruction-based multimodal image manipulation has recently made rapid progress. However, existing evaluation methods lack a systematic and human-aligned framework for assessing model performance on complex and creative editing tasks. To address this gap, we propose CREval, a fully automated question-answer (QA)-based evaluation pipeline that overcomes the incompleteness and poor interpretability of opaque Multimodal Large Language Models (MLLMs) scoring. Simultaneously, we introduce CREval-Bench, a comprehensive benchmark specifically designed for creative image manipulation under complex instructions. CREval-Bench covers three categories and nine creative dimensions, comprising over 800 editing samples and 13K evaluation queries. Leveraging this pipeline and benchmark, we systematically evaluate a diverse set of state-of-the-art open and closed-source models. The results reveal that while closed-source models generally outperform open-source ones on complex and creative tasks, all models still struggle to complete such edits effectively. In addition, user studies demonstrate strong consistency between CREval's automated metrics and human judgments. Therefore, CREval provides a reliable foundation for evaluating image editing models on complex and creative image manipulation tasks, and highlights key challenges and opportunities for future research.

Xi Chen, Haosen Yang, Sheng Jin et al.

Open-vocabulary segmentation poses significant challenges, as it requires segmenting and recognizing objects across an open set of categories in unconstrained environments. Building on the success of powerful vision-language (ViL) foundation models, such as CLIP, recent efforts sought to harness their zero-short capabilities to recognize unseen categories. Despite notable performance improvements, these models still encounter the critical issue of generating precise mask proposals for unseen categories and scenarios, resulting in inferior segmentation performance eventually. To address this challenge, we introduce a novel approach, FrozenSeg, designed to integrate spatial knowledge from a localization foundation model (e.g., SAM) and semantic knowledge extracted from a ViL model (e.g., CLIP), in a synergistic framework. Taking the ViL model's visual encoder as the feature backbone, we inject the space-aware feature into the learnable queries and CLIP features within the transformer decoder. In addition, we devise a mask proposal ensemble strategy for further improving the recall rate and mask quality. To fully exploit pre-trained knowledge while minimizing training overhead, we freeze both foundation models, focusing optimization efforts solely on a lightweight transformer decoder for mask proposal generation-the performance bottleneck. Extensive experiments demonstrate that FrozenSeg advances state-of-the-art results across various segmentation benchmarks, trained exclusively on COCO panoptic data, and tested in a zero-shot manner. Code is available at https://github.com/chenxi52/FrozenSeg.

Xi Chen, Hongxun Yao, Sicheng Zhao et al.

Source-free domain adaptation (SFDA) tackles the critical challenge of adapting source-pretrained models to unlabeled target domains without access to source data, overcoming data privacy and storage limitations in real-world applications. However, existing SFDA approaches struggle with the trade-off between perception field and computational efficiency in domain-invariant feature learning. Recently, Mamba has offered a promising solution through its selective scan mechanism, which enables long-range dependency modeling with linear complexity. However, the Visual Mamba (i.e., VMamba) remains limited in capturing channel-wise frequency characteristics critical for domain alignment and maintaining spatial robustness under significant domain shifts. To address these, we propose a framework called SfMamba to fully explore the stable dependency in source-free model transfer. SfMamba introduces Channel-wise Visual State-Space block that enables channel-sequence scanning for domain-invariant feature extraction. In addition, SfMamba involves a Semantic-Consistent Shuffle strategy that disrupts background patch sequences in 2D selective scan while preserving prediction consistency to mitigate error accumulation. Comprehensive evaluations across multiple benchmarks show that SfMamba achieves consistently stronger performance than existing methods while maintaining favorable parameter efficiency, offering a practical solution for SFDA. Our code is available at https://github.com/chenxi52/SfMamba.

Jing Jiang, Sicheng Zhao, Jiankun Zhu et al.

Unsupervised domain adaptation methods for panoramic semantic segmentation utilize real pinhole images or low-cost synthetic panoramic images to transfer segmentation models to real panoramic images. However, these methods struggle to understand the panoramic structure using only real pinhole images and lack real-world scene perception with only synthetic panoramic images. Therefore, in this paper, we propose a new task, Multi-source Domain Adaptation for Panoramic Semantic Segmentation (MSDA4PASS), which leverages both real pinhole and synthetic panoramic images to improve segmentation on unlabeled real panoramic images. There are two key issues in the MSDA4PASS task: (1) distortion gaps between the pinhole and panoramic domains -- panoramic images exhibit global and local distortions absent in pinhole images; (2) texture gaps between the source and target domains -- scenes and styles differ across domains. To address these two issues, we propose a novel framework, Deformation Transform Aligner for Panoramic Semantic Segmentation (DTA4PASS), which converts all pinhole images in the source domains into distorted images and aligns the source distorted and panoramic images with the target panoramic images. Specifically, DTA4PASS consists of two main components: Unpaired Semantic Morphing (USM) and Distortion Gating Alignment (DGA). First, in USM, the Dual-view Discriminator (DvD) assists in training the diffeomorphic deformation network at the image and pixel level, enabling the effective deformation transformation of pinhole images without paired panoramic views, alleviating distortion gaps. Second, DGA assigns pinhole-like (pin-like) and panoramic-like (pan-like) features to each image by gating, and aligns these two features through uncertainty estimation, reducing texture gaps.

Xi Chen, Haosen Yang, Huicong Zhang et al.

Source-free unsupervised domain adaptation (SFUDA) aims to enable the utilization of a pre-trained source model in an unlabeled target domain without access to source data. Self-training is a way to solve SFUDA, where confident target samples are iteratively selected as pseudo-labeled samples to guide target model learning. However, prior heuristic noisy pseudo-label filtering methods all involve introducing extra models, which are sensitive to model assumptions and may introduce additional errors or mislabeling. In this work, we propose a method called Uncertainty-aware Pseudo-label-filtering Adaptation (UPA) to efficiently address this issue in a coarse-to-fine manner. Specially, we first introduce a sample selection module named Adaptive Pseudo-label Selection (APS), which is responsible for filtering noisy pseudo labels. The APS utilizes a simple sample uncertainty estimation method by aggregating knowledge from neighboring samples and confident samples are selected as clean pseudo-labeled. Additionally, we incorporate Class-Aware Contrastive Learning (CACL) to mitigate the memorization of pseudo-label noise by learning robust pair-wise representation supervised by pseudo labels. Through extensive experiments conducted on three widely used benchmarks, we demonstrate that our proposed method achieves competitive performance on par with state-of-the-art SFUDA methods. Code is available at https://github.com/chenxi52/UPA.

Yufan Chen, Lizhen Wang, Qijing Li et al.

The ability to animate photo-realistic head avatars reconstructed from monocular portrait video sequences represents a crucial step in bridging the gap between the virtual and real worlds. Recent advancements in head avatar techniques, including explicit 3D morphable meshes (3DMM), point clouds, and neural implicit representation have been exploited for this ongoing research. However, 3DMM-based methods are constrained by their fixed topologies, point-based approaches suffer from a heavy training burden due to the extensive quantity of points involved, and the last ones suffer from limitations in deformation flexibility and rendering efficiency. In response to these challenges, we propose MonoGaussianAvatar (Monocular Gaussian Point-based Head Avatar), a novel approach that harnesses 3D Gaussian point representation coupled with a Gaussian deformation field to learn explicit head avatars from monocular portrait videos. We define our head avatars with Gaussian points characterized by adaptable shapes, enabling flexible topology. These points exhibit movement with a Gaussian deformation field in alignment with the target pose and expression of a person, facilitating efficient deformation. Additionally, the Gaussian points have controllable shape, size, color, and opacity combined with Gaussian splatting, allowing for efficient training and rendering. Experiments demonstrate the superior performance of our method, which achieves state-of-the-art results among previous methods.

Tingting Zheng, Kui Jiang, Hongxun Yao

Multi-Instance Learning (MIL) has shown impressive performance for histopathology whole slide image (WSI) analysis using bags or pseudo-bags. It involves instance sampling, feature representation, and decision-making. However, existing MIL-based technologies at least suffer from one or more of the following problems: 1) requiring high storage and intensive pre-processing for numerous instances (sampling); 2) potential over-fitting with limited knowledge to predict bag labels (feature representation); 3) pseudo-bag counts and prior biases affect model robustness and generalizability (decision-making). Inspired by clinical diagnostics, using the past sampling instances can facilitate the final WSI analysis, but it is barely explored in prior technologies. To break free these limitations, we integrate the dynamic instance sampling and reinforcement learning into a unified framework to improve the instance selection and feature aggregation, forming a novel Dynamic Policy Instance Selection (DPIS) scheme for better and more credible decision-making. Specifically, the measurement of feature distance and reward function are employed to boost continuous instance sampling. To alleviate the over-fitting, we explore the latent global relations among instances for more robust and discriminative feature representation while establishing reward and punishment mechanisms to correct biases in pseudo-bags using contrastive learning. These strategies form the final Dynamic Policy-Driven Adaptive Multi-Instance Learning (PAMIL) method for WSI tasks. Extensive experiments reveal that our PAMIL method outperforms the state-of-the-art by 3.8\% on CAMELYON16 and 4.4\% on TCGA lung cancer datasets.

Shengping Zhang, Xiaoyu Han, Weigang Zhang et al.

Image-based virtual try-on aims to transfer an in-shop clothing image to a person image. Most existing methods adopt a single global deformation to perform clothing warping directly, which lacks fine-grained modeling of in-shop clothing and leads to distorted clothing appearance. In addition, existing methods usually fail to generate limb details well because they are limited by the used clothing-agnostic person representation without referring to the limb textures of the person image. To address these problems, we propose Limb-aware Virtual Try-on Network named PL-VTON, which performs fine-grained clothing warping progressively and generates high-quality try-on results with realistic limb details. Specifically, we present Progressive Clothing Warping (PCW) that explicitly models the location and size of in-shop clothing and utilizes a two-stage alignment strategy to progressively align the in-shop clothing with the human body. Moreover, a novel gravity-aware loss that considers the fit of the person wearing clothing is adopted to better handle the clothing edges. Then, we design Person Parsing Estimator (PPE) with a non-limb target parsing map to semantically divide the person into various regions, which provides structural constraints on the human body and therefore alleviates texture bleeding between clothing and body regions. Finally, we introduce Limb-aware Texture Fusion (LTF) that focuses on generating realistic details in limb regions, where a coarse try-on result is first generated by fusing the warped clothing image with the person image, then limb textures are further fused with the coarse result under limb-aware guidance to refine limb details. Extensive experiments demonstrate that our PL-VTON outperforms the state-of-the-art methods both qualitatively and quantitatively.

Qinglin Liu, Shengping Zhang, Quanling Meng et al.

Human instance matting aims to estimate an alpha matte for each human instance in an image, which is extremely challenging and has rarely been studied so far. Despite some efforts to use instance segmentation to generate a trimap for each instance and apply trimap-based matting methods, the resulting alpha mattes are often inaccurate due to inaccurate segmentation. In addition, this approach is computationally inefficient due to multiple executions of the matting method. To address these problems, this paper proposes a novel End-to-End Human Instance Matting (E2E-HIM) framework for simultaneous multiple instance matting in a more efficient manner. Specifically, a general perception network first extracts image features and decodes instance contexts into latent codes. Then, a united guidance network exploits spatial attention and semantics embedding to generate united semantics guidance, which encodes the locations and semantic correspondences of all instances. Finally, an instance matting network decodes the image features and united semantics guidance to predict all instance-level alpha mattes. In addition, we construct a large-scale human instance matting dataset (HIM-100K) comprising over 100,000 human images with instance alpha matte labels. Experiments on HIM-100K demonstrate the proposed E2E-HIM outperforms the existing methods on human instance matting with 50% lower errors and 5X faster speed (6 instances in a 640X640 image). Experiments on the PPM-100, RWP-636, and P3M datasets demonstrate that E2E-HIM also achieves competitive performance on traditional human matting.

Jiankun Zhu, Sicheng Zhao, Jing Jiang et al.

Visual emotion recognition (VER), which aims at understanding humans' emotional reactions toward different visual stimuli, has attracted increasing attention. Given the subjective and ambiguous characteristics of emotion, annotating a reliable large-scale dataset is hard. For reducing reliance on data labeling, domain adaptation offers an alternative solution by adapting models trained on labeled source data to unlabeled target data. Conventional domain adaptation methods require access to source data. However, due to privacy concerns, source emotional data may be inaccessible. To address this issue, we propose an unexplored task: source-free domain adaptation (SFDA) for VER, which does not have access to source data during the adaptation process. To achieve this, we propose a novel framework termed Bridge then Begin Anew (BBA), which consists of two steps: domain-bridged model generation (DMG) and target-related model adaptation (TMA). First, the DMG bridges cross-domain gaps by generating an intermediate model, avoiding direct alignment between two VER datasets with significant differences. Then, the TMA begins training the target model anew to fit the target structure, avoiding the influence of source-specific knowledge. Extensive experiments are conducted on six SFDA settings for VER. The results demonstrate the effectiveness of BBA, which achieves remarkable performance gains compared with state-of-the-art SFDA methods and outperforms representative unsupervised domain adaptation approaches.

Zhaopan Xu, Pengfei Zhou, Jiaxin Ai et al.

Reasoning is an essential capacity for large language models (LLMs) to address complex tasks, where the identification of process errors is vital for improving this ability. Recently, process-level reward models (PRMs) were proposed to provide step-wise rewards that facilitate reinforcement learning and data production during training and guide LLMs toward correct steps during inference, thereby improving reasoning accuracy. However, existing benchmarks of PRMs are text-based and focus on error detection, neglecting other scenarios like reasoning search. To address this gap, we introduce MPBench, a comprehensive, multi-task, multimodal benchmark designed to systematically assess the effectiveness of PRMs in diverse scenarios. MPBench employs three evaluation paradigms, each targeting a specific role of PRMs in the reasoning process: (1) Step Correctness, which assesses the correctness of each intermediate reasoning step; (2) Answer Aggregation, which aggregates multiple solutions and selects the best one; and (3) Reasoning Process Search, which guides the search for optimal reasoning steps during inference. Through these paradigms, MPBench makes comprehensive evaluations and provides insights into the development of multimodal PRMs.

Zhaopan Xu, Pengfei Zhou, Weidong Tang et al.

Multimodal large language models (MLLMs) have achieved remarkable success in vision-language tasks, but their reliance on vast, internet-sourced data raises significant privacy and security concerns. Machine unlearning (MU) has emerged as a critical technique to address these issues, enabling the selective removal of targeted information from pre-trained models without costly retraining. However, the evaluation of MU for MLLMs remains inadequate. Existing benchmarks often lack a comprehensive scope, focusing narrowly on entities while overlooking the unlearning of broader visual concepts and the inherent semantic coupling between them. To bridge this gap, we introduce, PEBench, a novel benchmark designed to facilitate a thorough assessment of MU in MLLMs. PEBench features a fictitious dataset of personal entities and corresponding event scenes to evaluate unlearning across these distinct yet entangled concepts. We leverage this benchmark to evaluate five MU methods, revealing their unique strengths and weaknesses. Our findings show that unlearning one concept can unintentionally degrade performance on related concepts within the same image, a challenge we term cross-concept interference. Furthermore, we demonstrate the difficulty of unlearning person and event concepts simultaneously and propose an effective method to mitigate these conflicting objectives. The source code and benchmark are publicly available at https://pebench.github.io.

Xi Chen, Hongxun Yao, Zhaopan Xu et al.

Source-free active domain adaptation (SFADA) enhances knowledge transfer from a source model to an unlabeled target domain using limited manual labels selected via active learning. While recent domain adaptation studies have introduced Vision-and-Language (ViL) models to improve pseudo-label quality or feature alignment, they often treat ViL-based and data supervision as separate sources, lacking effective fusion. To overcome this limitation, we propose Dual Active learning with Multimodal (DAM) foundation model, a novel framework that integrates multimodal supervision from a ViL model to complement sparse human annotations, thereby forming a dual supervisory signal. DAM initializes stable ViL-guided targets and employs a bidirectional distillation mechanism to foster mutual knowledge exchange between the target model and the dual supervisions during iterative adaptation. Extensive experiments demonstrate that DAM consistently outperforms existing methods and sets a new state-of-the-art across multiple SFADA benchmarks and active learning strategies.

Shiyu Liu, Kui Jiang, Xianming Liu et al.

Audio-driven talking head video generation enhances user engagement in human-computer interaction. However, current methods frequently produce videos with motion blur and lip jitter, primarily due to their reliance on implicit modeling of audio-facial motion correlations--an approach lacking explicit articulatory priors (i.e., anatomical guidance for speech-related facial movements). To overcome this limitation, we propose HM-Talker, a novel framework for generating high-fidelity, temporally coherent talking heads. HM-Talker leverages a hybrid motion representation combining both implicit and explicit motion cues. Explicit cues use Action Units (AUs), anatomically defined facial muscle movements, alongside implicit features to minimize phoneme-viseme misalignment. Specifically, our Cross-Modal Disentanglement Module (CMDM) extracts complementary implicit/explicit motion features while predicting AUs directly from audio input aligned to visual cues. To mitigate identity-dependent biases in explicit features and enhance cross-subject generalization, we introduce the Hybrid Motion Modeling Module (HMMM). This module dynamically merges randomly paired implicit/explicit features, enforcing identity-agnostic learning. Together, these components enable robust lip synchronization across diverse identities, advancing personalized talking head synthesis. Extensive experiments demonstrate HM-Talker's superiority over state-of-the-art methods in visual quality and lip-sync accuracy.

Kui Jiang, Shiyu Liu, Junjun Jiang et al.

Audio-driven talking head generation holds significant potential for film production. While existing 3D methods have advanced motion modeling and content synthesis, they often produce rendering artifacts, such as motion blur, temporal jitter, and local penetration, due to limitations in representing stable, fine-grained motion fields. Through systematic analysis, we reformulate talking head generation into a unified framework comprising three steps: video preprocessing, motion representation, and rendering reconstruction. This framework underpins our proposed M2DAO-Talker, which addresses current limitations via multi-granular motion decoupling and alternating optimization. Specifically, we devise a novel 2D portrait preprocessing pipeline to extract frame-wise deformation control conditions (motion region segmentation masks, and camera parameters) to facilitate motion representation. To ameliorate motion modeling, we elaborate a multi-granular motion decoupling strategy, which independently models non-rigid (oral and facial) and rigid (head) motions for improved reconstruction accuracy. Meanwhile, a motion consistency constraint is developed to ensure head-torso kinematic consistency, thereby mitigating penetration artifacts caused by motion aliasing. In addition, an alternating optimization strategy is designed to iteratively refine facial and oral motion parameters, enabling more realistic video generation. Experiments across multiple datasets show that M2DAO-Talker achieves state-of-the-art performance, with the 2.43 dB PSNR improvement in generation quality and 0.64 gain in user-evaluated video realness versus TalkingGaussian while with 150 FPS inference speed. Our project homepage is https://m2dao-talker.github.io/M2DAO-Talk.github.io.

Huicong Zhang, Haozhe Xie, Hongxun Yao

Video deblurring relies on leveraging information from other frames in the video sequence to restore the blurred regions in the current frame. Mainstream approaches employ bidirectional feature propagation, spatio-temporal transformers, or a combination of both to extract information from the video sequence. However, limitations in memory and computational resources constraints the temporal window length of the spatio-temporal transformer, preventing the extraction of longer temporal contextual information from the video sequence. Additionally, bidirectional feature propagation is highly sensitive to inaccurate optical flow in blurry frames, leading to error accumulation during the propagation process. To address these issues, we propose \textbf{BSSTNet}, \textbf{B}lur-aware \textbf{S}patio-temporal \textbf{S}parse \textbf{T}ransformer Network. It introduces the blur map, which converts the originally dense attention into a sparse form, enabling a more extensive utilization of information throughout the entire video sequence. Specifically, BSSTNet (1) uses a longer temporal window in the transformer, leveraging information from more distant frames to restore the blurry pixels in the current frame. (2) introduces bidirectional feature propagation guided by blur maps, which reduces error accumulation caused by the blur frame. The experimental results demonstrate the proposed BSSTNet outperforms the state-of-the-art methods on the GoPro and DVD datasets.

Haosen Yang, Wenhao Wu, Lining Wang et al.

Temporal action proposal generation (TAPG) is a challenging task that aims to locate action instances in untrimmed videos with temporal boundaries. To evaluate the confidence of proposals, the existing works typically predict action score of proposals that are supervised by the temporal Intersection-over-Union (tIoU) between proposal and the ground-truth. In this paper, we innovatively propose a general auxiliary Background Constraint idea to further suppress low-quality proposals, by utilizing the background prediction score to restrict the confidence of proposals. In this way, the Background Constraint concept can be easily plug-and-played into existing TAPG methods (e.g., BMN, GTAD). From this perspective, we propose the Background Constraint Network (BCNet) to further take advantage of the rich information of action and background. Specifically, we introduce an Action-Background Interaction module for reliable confidence evaluation, which models the inconsistency between action and background by attention mechanisms at the frame and clip levels. Extensive experiments are conducted on two popular benchmarks, i.e., ActivityNet-1.3 and THUMOS14. The results demonstrate that our method outperforms state-of-the-art methods. Equipped with the existing action classifier, our method also achieves remarkable performance on the temporal action localization task.

Lining Wang, Haosen Yang, Wenhao Wu et al.

Transformer networks are effective at modeling long-range contextual information and have recently demonstrated exemplary performance in the natural language processing domain. Conventionally, the temporal action proposal generation (TAPG) task is divided into two main sub-tasks: boundary prediction and proposal confidence prediction, which rely on the frame-level dependencies and proposal-level relationships separately. To capture the dependencies at different levels of granularity, this paper intuitively presents a unified temporal action proposal generation framework with original Transformers, called TAPG Transformer, which consists of a Boundary Transformer and a Proposal Transformer. Specifically, the Boundary Transformer captures long-term temporal dependencies to predict precise boundary information and the Proposal Transformer learns the rich inter-proposal relationships for reliable confidence evaluation. Extensive experiments are conducted on two popular benchmarks: ActivityNet-1.3 and THUMOS14, and the results demonstrate that TAPG Transformer outperforms state-of-the-art methods. Equipped with the existing action classifier, our method achieves remarkable performance on the temporal action localization task. Codes and models will be available.

Haozhe Xie, Hongxun Yao, Shangchen Zhou et al.

Recently, several Space-Time Memory based networks have shown that the object cues (e.g. video frames as well as the segmented object masks) from the past frames are useful for segmenting objects in the current frame. However, these methods exploit the information from the memory by global-to-global matching between the current and past frames, which lead to mismatching to similar objects and high computational complexity. To address these problems, we propose a novel local-to-local matching solution for semi-supervised VOS, namely Regional Memory Network (RMNet). In RMNet, the precise regional memory is constructed by memorizing local regions where the target objects appear in the past frames. For the current query frame, the query regions are tracked and predicted based on the optical flow estimated from the previous frame. The proposed local-to-local matching effectively alleviates the ambiguity of similar objects in both memory and query frames, which allows the information to be passed from the regional memory to the query region efficiently and effectively. Experimental results indicate that the proposed RMNet performs favorably against state-of-the-art methods on the DAVIS and YouTube-VOS datasets.

Haozhe Xie, Hongxun Yao, Shengping Zhang et al.

Recovering the 3D shape of an object from single or multiple images with deep neural networks has been attracting increasing attention in the past few years. Mainstream works (e.g. 3D-R2N2) use recurrent neural networks (RNNs) to sequentially fuse feature maps of input images. However, RNN-based approaches are unable to produce consistent reconstruction results when given the same input images with different orders. Moreover, RNNs may forget important features from early input images due to long-term memory loss. To address these issues, we propose a novel framework for single-view and multi-view 3D object reconstruction, named Pix2Vox++. By using a well-designed encoder-decoder, it generates a coarse 3D volume from each input image. A multi-scale context-aware fusion module is then introduced to adaptively select high-quality reconstructions for different parts from all coarse 3D volumes to obtain a fused 3D volume. To further correct the wrongly recovered parts in the fused 3D volume, a refiner is adopted to generate the final output. Experimental results on the ShapeNet, Pix3D, and Things3D benchmarks show that Pix2Vox++ performs favorably against state-of-the-art methods in terms of both accuracy and efficiency.

Haozhe Xie, Hongxun Yao, Shangchen Zhou et al.

Estimating the complete 3D point cloud from an incomplete one is a key problem in many vision and robotics applications. Mainstream methods (e.g., PCN and TopNet) use Multi-layer Perceptrons (MLPs) to directly process point clouds, which may cause the loss of details because the structural and context of point clouds are not fully considered. To solve this problem, we introduce 3D grids as intermediate representations to regularize unordered point clouds. We therefore propose a novel Gridding Residual Network (GRNet) for point cloud completion. In particular, we devise two novel differentiable layers, named Gridding and Gridding Reverse, to convert between point clouds and 3D grids without losing structural information. We also present the differentiable Cubic Feature Sampling layer to extract features of neighboring points, which preserves context information. In addition, we design a new loss function, namely Gridding Loss, to calculate the L1 distance between the 3D grids of the predicted and ground truth point clouds, which is helpful to recover details. Experimental results indicate that the proposed GRNet performs favorably against state-of-the-art methods on the ShapeNet, Completion3D, and KITTI benchmarks.

Xinjie Feng, Hongxun Yao, Yuankai Qi et al.

Scene text recognition with arbitrary shape is very challenging due to large variations in text shapes, fonts, colors, backgrounds, etc. Most state-of-the-art algorithms rectify the input image into the normalized image, then treat the recognition as a sequence prediction task. The bottleneck of such methods is the rectification, which will cause errors due to distortion perspective. In this paper, we find that the rectification is completely unnecessary. What all we need is the spatial attention. We therefore propose a simple but extremely effective scene text recognition method based on transformer [50]. Different from previous transformer based models [56,34], which just use the decoder of the transformer to decode the convolutional attention, the proposed method use a convolutional feature maps as word embedding input into transformer. In such a way, our method is able to make full use of the powerful attention mechanism of the transformer. Extensive experimental results show that the proposed method significantly outperforms state-of-the-art methods by a very large margin on both regular and irregular text datasets. On one of the most challenging CUTE dataset whose state-of-the-art prediction accuracy is 89.6%, our method achieves 99.3%, which is a pretty surprising result. We will release our source code and believe that our method will be a new benchmark of scene text recognition with arbitrary shapes.

Shuo Yang, Wei Yu, Ying Zheng et al.

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

Sheng Jin, Shangchen Zhou, Yao Liu et al.

Deep hashing methods have been proved to be effective and efficient for large-scale Web media search. The success of these data-driven methods largely depends on collecting sufficient labeled data, which is usually a crucial limitation in practical cases. The current solutions to this issue utilize Generative Adversarial Network (GAN) to augment data in semi-supervised learning. However, existing GAN-based methods treat image generations and hashing learning as two isolated processes, leading to generation ineffectiveness. Besides, most works fail to exploit the semantic information in unlabeled data. In this paper, we propose a novel Semi-supervised Self-pace Adversarial Hashing method, named SSAH to solve the above problems in a unified framework. The SSAH method consists of an adversarial network (A-Net) and a hashing network (H-Net). To improve the quality of generative images, first, the A-Net learns hard samples with multi-scale occlusions and multi-angle rotated deformations which compete against the learning of accurate hashing codes. Second, we design a novel self-paced hard generation policy to gradually increase the hashing difficulty of generated samples. To make use of the semantic information in unlabeled ones, we propose a semi-supervised consistent loss. The experimental results show that our method can significantly improve state-of-the-art models on both the widely-used hashing datasets and fine-grained datasets.

Haozhe Xie, Hongxun Yao, Shangchen Zhou et al.

Inferring the 3D shape of an object from an RGB image has shown impressive results, however, existing methods rely primarily on recognizing the most similar 3D model from the training set to solve the problem. These methods suffer from poor generalization and may lead to low-quality reconstructions for unseen objects. Nowadays, stereo cameras are pervasive in emerging devices such as dual-lens smartphones and robots, which enables the use of the two-view nature of stereo images to explore the 3D structure and thus improve the reconstruction performance. In this paper, we propose a new deep learning framework for reconstructing the 3D shape of an object from a pair of stereo images, which reasons about the 3D structure of the object by taking bidirectional disparities and feature correspondences between the two views into account. Besides, we present a large-scale synthetic benchmarking dataset, namely StereoShapeNet, containing 1,052,976 pairs of stereo images rendered from ShapeNet along with the corresponding bidirectional depth and disparity maps. Experimental results on the StereoShapeNet benchmark demonstrate that the proposed framework outperforms the state-of-the-art methods.

Ying Zheng, Hongxun Yao, Xiaoshuai Sun et al.

Sketch recognition remains a significant challenge due to the limited training data and the substantial intra-class variance of freehand sketches for the same object. Conventional methods for this task often rely on the availability of the temporal order of sketch strokes, additional cues acquired from different modalities and supervised augmentation of sketch datasets with real images, which also limit the applicability and feasibility of these methods in real scenarios. In this paper, we propose a novel sketch-specific data augmentation (SSDA) method that leverages the quantity and quality of the sketches automatically. From the aspect of quantity, we introduce a Bezier pivot based deformation (BPD) strategy to enrich the training data. Towards quality improvement, we present a mean stroke reconstruction (MSR) approach to generate a set of novel types of sketches with smaller intra-class variances. Both of these solutions are unrestricted from any multi-source data and temporal cues of sketches. Furthermore, we show that some recent deep convolutional neural network models that are trained on generic classes of real images can be better choices than most of the elaborate architectures that are designed explicitly for sketch recognition. As SSDA can be integrated with any convolutional neural networks, it has a distinct advantage over the existing methods. Our extensive experimental evaluations demonstrate that the proposed method achieves the state-of-the-art results (84.27%) on the TU-Berlin dataset, outperforming the human performance by a remarkable 11.17% increase. Finally, more experiments show the practical value of our approach for the task of sketch-based image retrieval.

Ying Zheng, Hongxun Yao, Xiaoshuai Sun

In this paper, we propose a novel deep framework for part-level semantic parsing of freehand sketches, which makes three main contributions that are experimentally shown to have substantial practical merit. First, we propose a homogeneous transformation method to address the problem of domain adaptation. For the task of sketch parsing, there is no available data of labeled freehand sketches that can be directly used for model training. An alternative solution is to learn from datasets of real image parsing, while the domain adaptation is an inevitable problem. Unlike existing methods that utilize the edge maps of real images to approximate freehand sketches, the proposed homogeneous transformation method transforms the data from domains of real images and freehand sketches into a homogeneous space to minimize the semantic gap. Second, we design a soft-weighted loss function as guidance for the training process, which gives attention to both the ambiguous label boundary and class imbalance. Third, we present a staged learning strategy to improve the parsing performance of the trained model, which takes advantage of the shared information and specific characteristic from different sketch categories. Extensive experimental results demonstrate the effectiveness of the above three methods. Specifically, to evaluate the generalization ability of our homogeneous transformation method, additional experiments for the task of sketch-based image retrieval are conducted on the QMUL FG-SBIR dataset. Finally, by integrating the proposed three methods into a unified framework of deep semantic sketch parsing (DeepSSP), we achieve the state-of-the-art on the public SketchParse dataset.

Haozhe Xie, Hongxun Yao, Xiaoshuai Sun et al.

Recovering the 3D representation of an object from single-view or multi-view RGB images by deep neural networks has attracted increasing attention in the past few years. Several mainstream works (e.g., 3D-R2N2) use recurrent neural networks (RNNs) to fuse multiple feature maps extracted from input images sequentially. However, when given the same set of input images with different orders, RNN-based approaches are unable to produce consistent reconstruction results. Moreover, due to long-term memory loss, RNNs cannot fully exploit input images to refine reconstruction results. To solve these problems, we propose a novel framework for single-view and multi-view 3D reconstruction, named Pix2Vox. By using a well-designed encoder-decoder, it generates a coarse 3D volume from each input image. Then, a context-aware fusion module is introduced to adaptively select high-quality reconstructions for each part (e.g., table legs) from different coarse 3D volumes to obtain a fused 3D volume. Finally, a refiner further refines the fused 3D volume to generate the final output. Experimental results on the ShapeNet and Pix3D benchmarks indicate that the proposed Pix2Vox outperforms state-of-the-arts by a large margin. Furthermore, the proposed method is 24 times faster than 3D-R2N2 in terms of backward inference time. The experiments on ShapeNet unseen 3D categories have shown the superior generalization abilities of our method.

Sheng Jin, Hongxun Yao, Xiaoshuai Sun et al.

In recent years, hashing methods have been proved to be effective and efficient for the large-scale Web media search. However, the existing general hashing methods have limited discriminative power for describing fine-grained objects that share similar overall appearance but have subtle difference. To solve this problem, we for the first time introduce the attention mechanism to the learning of fine-grained hashing codes. Specifically, we propose a novel deep hashing model, named deep saliency hashing (DSaH), which automatically mines salient regions and learns semantic-preserving hashing codes simultaneously. DSaH is a two-step end-to-end model consisting of an attention network and a hashing network. Our loss function contains three basic components, including the semantic loss, the saliency loss, and the quantization loss. As the core of DSaH, the saliency loss guides the attention network to mine discriminative regions from pairs of images. We conduct extensive experiments on both fine-grained and general retrieval datasets for performance evaluation. Experimental results on fine-grained datasets, including Oxford Flowers-17, Stanford Dogs-120, and CUB Bird demonstrate that our DSaH performs the best for fine-grained retrieval task and beats the strongest competitor (DTQ) by approximately 10% on both Stanford Dogs-120 and CUB Bird. DSaH is also comparable to several state-of-the-art hashing methods on general datasets, including CIFAR-10 and NUS-WIDE.

Wei Yu, Kuiyuan Yang, Yalong Bai et al.

Convolutional Neural Networks (CNNs) have achieved comparable error rates to well-trained human on ILSVRC2014 image classification task. To achieve better performance, the complexity of CNNs is continually increasing with deeper and bigger architectures. Though CNNs achieved promising external classification behavior, understanding of their internal work mechanism is still limited. In this work, we attempt to understand the internal work mechanism of CNNs by probing the internal representations in two comprehensive aspects, i.e., visualizing patches in the representation spaces constructed by different layers, and visualizing visual information kept in each layer. We further compare CNNs with different depths and show the advantages brought by deeper architecture.