Jia Li

Yifan Zhao, Jia Li, Xiaowu Chen et al. · pku

Fine-grained visual recognition is to classify objects with visually similar appearances into subcategories, which has made great progress with the development of deep CNNs. However, handling subtle differences between different subcategories still remains a challenge. In this paper, we propose to solve this issue in one unified framework from two aspects, i.e., constructing feature-level interrelationships, and capturing part-level discriminative features. This framework, namely PArt-guided Relational Transformers (PART), is proposed to learn the discriminative part features with an automatic part discovery module, and to explore the intrinsic correlations with a feature transformation module by adapting the Transformer models from the field of natural language processing. The part discovery module efficiently discovers the discriminative regions which are highly-corresponded to the gradient descent procedure. Then the second feature transformation module builds correlations within the global embedding and multiple part embedding, enhancing spatial interactions among semantic pixels. Moreover, our proposed approach does not rely on additional part branches in the inference time and reaches state-of-the-art performance on 3 widely-used fine-grained object recognition benchmarks. Experimental results and explainable visualizations demonstrate the effectiveness of our proposed approach. The code can be found at https://github.com/iCVTEAM/PART.

Jianheng Tang, Fengrui Hua, Ziqi Gao et al.

With a long history of traditional Graph Anomaly Detection (GAD) algorithms and recently popular Graph Neural Networks (GNNs), it is still not clear (1) how they perform under a standard comprehensive setting, (2) whether GNNs can outperform traditional algorithms such as tree ensembles, and (3) how about their efficiency on large-scale graphs. In response, we introduce GADBench -- a benchmark tool dedicated to supervised anomalous node detection in static graphs. GADBench facilitates a detailed comparison across 29 distinct models on ten real-world GAD datasets, encompassing thousands to millions ($\sim$6M) nodes. Our main finding is that tree ensembles with simple neighborhood aggregation can outperform the latest GNNs tailored for the GAD task. We shed light on the current progress of GAD, setting a robust groundwork for subsequent investigations in this domain. GADBench is open-sourced at https://github.com/squareRoot3/GADBench.

Yifan Zhao, Tong Zhang, Jia Li et al. · pku

Few-shot learning aims to recognize novel queries with limited support samples by learning from base knowledge. Recent progress in this setting assumes that the base knowledge and novel query samples are distributed in the same domains, which are usually infeasible for realistic applications. Toward this issue, we propose to address the cross-domain few-shot learning problem where only extremely few samples are available in target domains. Under this realistic setting, we focus on the fast adaptation capability of meta-learners by proposing an effective dual adaptive representation alignment approach. In our approach, a prototypical feature alignment is first proposed to recalibrate support instances as prototypes and reproject these prototypes with a differentiable closed-form solution. Therefore feature spaces of learned knowledge can be adaptively transformed to query spaces by the cross-instance and cross-prototype relations. Besides the feature alignment, we further present a normalized distribution alignment module, which exploits prior statistics of query samples for solving the covariant shifts among the support and query samples. With these two modules, a progressive meta-learning framework is constructed to perform the fast adaptation with extremely few-shot samples while maintaining its generalization capabilities. Experimental evidence demonstrates our approach achieves new state-of-the-art results on 4 CDFSL benchmarks and 4 fine-grained cross-domain benchmarks.

Yifan Zhao, Jia Li, Yonghong Tian · pku

Fine-grained visual parsing, including fine-grained part segmentation and fine-grained object recognition, has attracted considerable critical attention due to its importance in many real-world applications, e.g., agriculture, remote sensing, and space technologies. Predominant research efforts tackle these fine-grained sub-tasks following different paradigms, while the inherent relations between these tasks are neglected. Moreover, given most of the research remains fragmented, we conduct an in-depth study of the advanced work from a new perspective of learning the part relationship. In this perspective, we first consolidate recent research and benchmark syntheses with new taxonomies. Based on this consolidation, we revisit the universal challenges in fine-grained part segmentation and recognition tasks and propose new solutions by part relationship learning for these important challenges. Furthermore, we conclude several promising lines of research in fine-grained visual parsing for future research.

Shiming Ge, Shengwei Zhao, Chenyu Li et al.

Face recognition in the wild is now advancing towards light-weight models, fast inference speed and resolution-adapted capability. In this paper, we propose a bridge distillation approach to turn a complex face model pretrained on private high-resolution faces into a light-weight one for low-resolution face recognition. In our approach, such a cross-dataset resolution-adapted knowledge transfer problem is solved via two-step distillation. In the first step, we conduct cross-dataset distillation to transfer the prior knowledge from private high-resolution faces to public high-resolution faces and generate compact and discriminative features. In the second step, the resolution-adapted distillation is conducted to further transfer the prior knowledge to synthetic low-resolution faces via multi-task learning. By learning low-resolution face representations and mimicking the adapted high-resolution knowledge, a light-weight student model can be constructed with high efficiency and promising accuracy in recognizing low-resolution faces. Experimental results show that the student model performs impressively in recognizing low-resolution faces with only 0.21M parameters and 0.057MB memory. Meanwhile, its speed reaches up to 14,705, ~934 and 763 faces per second on GPU, CPU and mobile phone, respectively.

Yunshan Qi, Jia Li, Yifan Zhao et al. · pku

Neural Radiance Fields (NeRF) achieves impressive novel view rendering performance by learning implicit 3D representation from sparse view images. However, it is difficult to reconstruct a sharp NeRF from blurry input that often occurs in the wild. To solve this problem, we propose a novel Efficient Event-Enhanced NeRF (E$^{3}$NeRF), reconstructing sharp NeRF by utilizing both blurry images and corresponding event streams. A blur rendering loss and an event rendering loss are introduced, which guide the NeRF training via modeling the physical image motion blur process and event generation process, respectively. To improve the efficiency of the framework, we further leverage the latent spatial-temporal blur information in the event stream to evenly distribute training over temporal blur and focus training on spatial blur. Moreover, a camera pose estimation framework for real-world data is built with the guidance of the events, generalizing the method to more practical applications. Compared to previous image-based and event-based NeRF works, our framework makes more profound use of the internal relationship between events and images. Extensive experiments on both synthetic data and real-world data demonstrate that E\textsuperscript{3}NeRF can effectively learn a sharp NeRF from blurry images, especially for high-speed non-uniform motion and low-light scenes.

Junjie Wu, Changqun Xia, Tianshu Yu et al.

Image-based salient object detection (ISOD) in 360° scenarios is significant for understanding and applying panoramic information. However, research on 360° ISOD has not been widely explored due to the lack of large, complex, high-resolution, and well-labeled datasets. Towards this end, we construct a large scale 360° ISOD dataset with object-level pixel-wise annotation on equirectangular projection (ERP), which contains rich panoramic scenes with not less than 2K resolution and is the largest dataset for 360° ISOD by far to our best knowledge. By observing the data, we find current methods face three significant challenges in panoramic scenarios: diverse distortion degrees, discontinuous edge effects and changeable object scales. Inspired by humans' observing process, we propose a view-aware salient object detection method based on a Sample Adaptive View Transformer (SAVT) module with two sub-modules to mitigate these issues. Specifically, the sub-module View Transformer (VT) contains three transform branches based on different kinds of transformations to learn various features under different views and heighten the model's feature toleration of distortion, edge effects and object scales. Moreover, the sub-module Sample Adaptive Fusion (SAF) is to adjust the weights of different transform branches based on various sample features and make transformed enhanced features fuse more appropriately. The benchmark results of 20 state-of-the-art ISOD methods reveal the constructed dataset is very challenging. Moreover, exhaustive experiments verify the proposed approach is practical and outperforms the state-of-the-art methods.

Yin Chen, Jia Li, Yu Zhang et al.

Dynamic facial expression recognition (DFER) infers emotions from the temporal evolution of expressions, unlike static facial expression recognition (SFER), which relies solely on a single snapshot. This temporal analysis provides richer information and promises greater recognition capability. However, current DFER methods often exhibit unsatisfied performance largely due to fewer training samples compared to SFER. Given the inherent correlation between static and dynamic expressions, we hypothesize that leveraging the abundant SFER data can enhance DFER. To this end, we propose Static-for-Dynamic (S4D), a unified dual-modal learning framework that integrates SFER data as a complementary resource for DFER. Specifically, S4D employs dual-modal self-supervised pre-training on facial images and videos using a shared Vision Transformer (ViT) encoder-decoder architecture, yielding improved spatiotemporal representations. The pre-trained encoder is then fine-tuned on static and dynamic expression datasets in a multi-task learning setup to facilitate emotional information interaction. Unfortunately, vanilla multi-task learning in our study results in negative transfer. To address this, we propose an innovative Mixture of Adapter Experts (MoAE) module that facilitates task-specific knowledge acquisition while effectively extracting shared knowledge from both static and dynamic expression data. Extensive experiments demonstrate that S4D achieves a deeper understanding of DFER, setting new state-of-the-art performance on FERV39K, MAFW, and DFEW benchmarks, with weighted average recall (WAR) of 53.65\%, 58.44\%, and 76.68\%, respectively. Additionally, a systematic correlation analysis between SFER and DFER tasks is presented, which further elucidates the potential benefits of leveraging SFER.

Jian Xiao, Zijie Song, Jialong Hu et al.

Recent progress in text-video retrieval has been largely driven by contrastive learning. However, existing methods often overlook the effect of the modality gap, which causes anchor representations to undergo in-place optimization (i.e., optimization tension) that limits their alignment capacity. Moreover, noisy hard negatives further distort the semantics of anchors. To address these issues, we propose GARE, a Gap-Aware Retrieval framework that introduces a learnable, pair-specific increment $Δ_{ij}$ between text $t_i$ and video $v_j$, redistributing gradients to relieve optimization tension and absorb noise. We derive $Δ_{ij}$ via a multivariate first-order Taylor expansion of the InfoNCE loss under a trust-region constraint, showing that it guides updates along locally consistent descent directions. A lightweight neural module conditioned on the semantic gap couples increments across batches for structure-aware correction. Furthermore, we regularize $Δ$ through a variational information bottleneck with relaxed compression, enhancing stability and semantic consistency. Experiments on four benchmarks demonstrate that GARE consistently improves alignment accuracy and robustness, validating the effectiveness of gap-aware tension mitigation. Code is available at https://github.com/musicman217/GARE-text-video-retrieval.

Yifan Zhao, Jia Li, Zeyin Song et al. · pku

Depicting novel classes with language descriptions by observing few-shot samples is inherent in human-learning systems. This lifelong learning capability helps to distinguish new knowledge from old ones through the increase of open-world learning, namely Few-Shot Class-Incremental Learning (FSCIL). Existing works to solve this problem mainly rely on the careful tuning of visual encoders, which shows an evident trade-off between the base knowledge and incremental ones. Motivated by human learning systems, we propose a new Language-inspired Relation Transfer (LRT) paradigm to understand objects by joint visual clues and text depictions, composed of two major steps. We first transfer the pretrained text knowledge to the visual domains by proposing a graph relation transformation module and then fuse the visual and language embedding by a text-vision prototypical fusion module. Second, to mitigate the domain gap caused by visual finetuning, we propose context prompt learning for fast domain alignment and imagined contrastive learning to alleviate the insufficient text data during alignment. With collaborative learning of domain alignments and text-image transfer, our proposed LRT outperforms the state-of-the-art models by over $13\%$ and $7\%$ on the final session of mini-ImageNet and CIFAR-100 FSCIL benchmarks.

Ruitao Wu, Yifan Zhao, Jia Li · pku

Class-Incremental Semantic Segmentation (CISS) requires continuous learning of newly introduced classes while retaining knowledge of past classes. By abstracting mainstream methods into two stages (visual feature extraction and prototype-feature matching), we identify a more fundamental challenge termed catastrophic semantic entanglement. This phenomenon involves Prototype-Feature Entanglement caused by semantic misalignment during the incremental process, and Background-Increment Entanglement due to dynamic data evolution. Existing techniques, which rely on visual feature learning without sufficient cues to distinguish targets, introduce significant noise and errors. To address these issues, we introduce a Language-inspired Bootstrapped Disentanglement framework (LBD). We leverage the prior class semantics of pre-trained visual-language models (e.g., CLIP) to guide the model in autonomously disentangling features through Language-guided Prototypical Disentanglement and Manifold Mutual Background Disentanglement. The former guides the disentangling of new prototypes by treating hand-crafted text features as topological templates, while the latter employs multiple learnable prototypes and mask-pooling-based supervision for background-incremental class disentanglement. By incorporating soft prompt tuning and encoder adaptation modifications, we further bridge the capability gap of CLIP between dense and sparse tasks, achieving state-of-the-art performance on both Pascal VOC and ADE20k, particularly in multi-step scenarios.

Yangche Yu, Yin Chen, Jia Li et al.

Accurate engagement estimation is essential for adaptive human-computer interaction systems, yet robust deployment is hindered by poor generalizability across diverse domains and challenges in modeling complex interaction dynamics.To tackle these issues, we propose DAPA (Domain-Adaptive Parallel Attention), a novel framework for generalizable conversational engagement modeling. DAPA introduces a Domain Prompting mechanism by prepending learnable domain-specific vectors to the input, explicitly conditioning the model on the data's origin to facilitate domain-aware adaptation while preserving generalizable engagement representations. To capture interactional synchrony, the framework also incorporates a Parallel Cross-Attention module that explicitly aligns reactive (forward BiLSTM) and anticipatory (backward BiLSTM) states between participants.Extensive experiments demonstrate that DAPA establishes a new state-of-the-art performance on several cross-cultural and cross-linguistic benchmarks, notably achieving an absolute improvement of 0.45 in Concordance Correlation Coefficient (CCC) over a strong baseline on the NoXi-J test set. The superiority of our method was also confirmed by winning the first place in the Multi-Domain Engagement Estimation Challenge at MultiMediate'25.

Liyi Cai, Yijie Ren, Yitong Zhang et al. · tsinghua

Software automation has long been a central goal of software engineering, striving for software development that proceeds without human intervention. Recent efforts have leveraged Artificial Intelligence (AI) to advance software automation with notable progress. However, current AI functions primarily as assistants to human developers, leaving software development still dependent on explicit human intervention. This raises a fundamental question: Can AI move beyond its role as an assistant to become a core component of software, thereby enabling genuine software automation? To investigate this vision, we introduce AI-Driven Self-Evolving Software, a new form of software that evolves continuously through direct interaction with users. We demonstrate the feasibility of this idea with a lightweight prototype built on a multi-agent architecture that autonomously interprets user requirements, generates and validates code, and integrates new functionalities. Case studies across multiple representative scenarios show that the prototype can reliably construct and reuse functionality, providing early evidence that such software systems can scale to more sophisticated applications and pave the way toward truly automated software development. We make code and cases in this work publicly available at https://anonymous.4open.science/r/live-software.

Yunshan Qi, Lin Zhu, Yifan Zhao et al.

Neural Radiance Fields (NeRF) achieves impressive 3D representation learning and novel view synthesis results with high-quality multi-view images as input. However, motion blur in images often occurs in low-light and high-speed motion scenes, which significantly degrades the reconstruction quality of NeRF. Previous deblurring NeRF methods struggle to estimate pose and lighting changes during the exposure time, making them unable to accurately model the motion blur. The bio-inspired event camera measuring intensity changes with high temporal resolution makes up this information deficiency. In this paper, we propose Event-driven Bundle Adjustment for Deblurring Neural Radiance Fields (EBAD-NeRF) to jointly optimize the learnable poses and NeRF parameters by leveraging the hybrid event-RGB data. An intensity-change-metric event loss and a photo-metric blur loss are introduced to strengthen the explicit modeling of camera motion blur. Experiments on both synthetic and real-captured data demonstrate that EBAD-NeRF can obtain accurate camera trajectory during the exposure time and learn a sharper 3D representations compared to prior works.

Vincent Pisztora, Jia Li

In this paper we propose a method for the optimal allocation of observations between an intrinsically explainable glass box model and a black box model. An optimal allocation being defined as one which, for any given explainability level (i.e. the proportion of observations for which the explainable model is the prediction function), maximizes the performance of the ensemble on the underlying task, and maximizes performance of the explainable model on the observations allocated to it, subject to the maximal ensemble performance condition. The proposed method is shown to produce such explainability optimal allocations on a benchmark suite of tabular datasets across a variety of explainable and black box model types. These learned allocations are found to consistently maintain ensemble performance at very high explainability levels (explaining $74\%$ of observations on average), and in some cases even outperforming both the component explainable and black box models while improving explainability.

Mingcan Ma, Changqun Xia, Chenxi Xie et al.

Salient object detection requires a comprehensive and scalable receptive field to locate the visually significant objects in the image. Recently, the emergence of visual transformers and multi-branch modules has significantly enhanced the ability of neural networks to perceive objects at different scales. However, compared to the traditional backbone, the calculation process of transformers is time-consuming. Moreover, different branches of the multi-branch modules could cause the same error back propagation in each training iteration, which is not conducive to extracting discriminative features. To solve these problems, we propose a bilateral network based on transformer and CNN to efficiently broaden local details and global semantic information simultaneously. Besides, a Multi-Head Boosting (MHB) strategy is proposed to enhance the specificity of different network branches. By calculating the errors of different prediction heads, each branch can separately pay more attention to the pixels that other branches predict incorrectly. Moreover, Unlike multi-path parallel training, MHB randomly selects one branch each time for gradient back propagation in a boosting way. Additionally, an Attention Feature Fusion Module (AF) is proposed to fuse two types of features according to respective characteristics. Comprehensive experiments on five benchmark datasets demonstrate that the proposed method can achieve a significant performance improvement compared with the state-of-the-art methods.

Jia Li, Huaibo Huang, Xiaofei Jia et al.

Blind face restoration (BFR) is a challenging problem because of the uncertainty of the degradation patterns. This paper proposes a Restoration with Memorized Modulation (RMM) framework for universal BFR in diverse degraded scenes and heterogeneous domains. We apply random noise as well as unsupervised wavelet memory to adaptively modulate the face-enhancement generator, considering attentional denormalization in both layer and instance levels. Specifically, in the training stage, the low-level spatial feature embedding, the wavelet memory embedding obtained by wavelet transformation of the high-resolution image, as well as the disentangled high-level noise embeddings are integrated, with the guidance of attentional maps generated from layer normalization, instance normalization, and the original feature map. These three embeddings are respectively associated with the spatial content, high-frequency texture details, and a learnable universal prior against other blind image degradation patterns. We store the spatial feature of the low-resolution image and the corresponding wavelet style code as key and value in the memory unit, respectively. In the test stage, the wavelet memory value whose corresponding spatial key is the most matching with that of the inferred image is retrieved to modulate the generator. Moreover, the universal prior learned from the random noise has been memorized by training the modulation network. Experimental results show the superiority of the proposed method compared with the state-of-the-art methods, and a good generalization in the wild.

Yongjun Chen, Jia Li, Chenghao Liu et al.

Traditional approaches to next item and next basket recommendation typically extract users' interests based on their past interactions and associated static contextual information (e.g. a user id or item category). However, extracted interests can be inaccurate and become obsolete. Dynamic attributes, such as user income changes, item price changes (etc.), change over time. Such dynamics can intrinsically reflect the evolution of users' interests. We argue that modeling such dynamic attributes can boost recommendation performance. However, properly integrating them into user interest models is challenging since attribute dynamics can be diverse such as time-interval aware, periodic patterns (etc.), and they represent users' behaviors from different perspectives, which can happen asynchronously with interactions. Besides dynamic attributes, items in each basket contain complex interdependencies which might be beneficial but nontrivial to effectively capture. To address these challenges, we propose a novel Attentive network to model Dynamic attributes (named AnDa). AnDa separately encodes dynamic attributes and basket item sequences. We design a periodic aware encoder to allow the model to capture various temporal patterns from dynamic attributes. To effectively learn useful item relationships, intra-basket attention module is proposed. Experimental results on three real-world datasets demonstrate that our method consistently outperforms the state-of-the-art.

Zhongxing Ma, Yifan Zhao, Jia Li

Person Re-Identification (Re-Id) in occlusion scenarios is a challenging problem because a pedestrian can be partially occluded. The use of local information for feature extraction and matching is still necessary. Therefore, we propose a Pose-guided inter-and intra-part relational transformer (Pirt) for occluded person Re-Id, which builds part-aware long-term correlations by introducing transformers. In our framework, we firstly develop a pose-guided feature extraction module with regional grouping and mask construction for robust feature representations. The positions of a pedestrian in the image under surveillance scenarios are relatively fixed, hence we propose an intra-part and inter-part relational transformer. The intra-part module creates local relations with mask-guided features, while the inter-part relationship builds correlations with transformers, to develop cross relationships between part nodes. With the collaborative learning inter- and intra-part relationships, experiments reveal that our proposed Pirt model achieves a new state of the art on the public occluded dataset, and further extensions on standard non-occluded person Re-Id datasets also reveal our comparable performances.

Jia Li, Tomas Yu, Da-Cheng Juan et al.

Recent studies have indicated that Graph Convolutional Networks (GCNs) act as a \emph{low pass} filter in spectral domain and encode smoothed node representations. In this paper, we consider their opposite, namely Graph Deconvolutional Networks (GDNs) that reconstruct graph signals from smoothed node representations. We motivate the design of Graph Deconvolutional Networks via a combination of inverse filters in spectral domain and de-noising layers in wavelet domain, as the inverse operation results in a \emph{high pass} filter and may amplify the noise. Based on the proposed GDN, we further propose a graph autoencoder framework that first encodes smoothed graph representations with GCN and then decodes accurate graph signals with GDN. We demonstrate the effectiveness of the proposed method on several tasks including unsupervised graph-level representation , social recommendation and graph generation

Guangyao Chen, Limeng Qiao, Yemin Shi et al.

Open set recognition is an emerging research area that aims to simultaneously classify samples from predefined classes and identify the rest as 'unknown'. In this process, one of the key challenges is to reduce the risk of generalizing the inherent characteristics of numerous unknown samples learned from a small amount of known data. In this paper, we propose a new concept, Reciprocal Point, which is the potential representation of the extra-class space corresponding to each known category. The sample can be classified to known or unknown by the otherness with reciprocal points. To tackle the open set problem, we offer a novel open space risk regularization term. Based on the bounded space constructed by reciprocal points, the risk of unknown is reduced through multi-category interaction. The novel learning framework called Reciprocal Point Learning (RPL), which can indirectly introduce the unknown information into the learner with only known classes, so as to learn more compact and discriminative representations. Moreover, we further construct a new large-scale challenging aircraft dataset for open set recognition: Aircraft 300 (Air-300). Extensive experiments on multiple benchmark datasets indicate that our framework is significantly superior to other existing approaches and achieves state-of-the-art performance on standard open set benchmarks.

Xin Guo, Yifan Zhao, Jia Li

Close your eyes and listen to music, one can easily imagine an actor dancing rhythmically along with the music. These dance movements are usually made up of dance movements you have seen before. In this paper, we propose to reproduce such an inherent capability of the human-being within a computer vision system. The proposed system consists of three modules. To explore the relationship between music and dance movements, we propose a cross-modal alignment module that focuses on dancing video clips, accompanied on pre-designed music, to learn a system that can judge the consistency between the visual features of pose sequences and the acoustic features of music. The learned model is then used in the imagination module to select a pose sequence for the given music. Such pose sequence selected from the music, however, is usually discontinuous. To solve this problem, in the spatial-temporal alignment module we develop a spatial alignment algorithm based on the tendency and periodicity of dance movements to predict dance movements between discontinuous fragments. In addition, the selected pose sequence is often misaligned with the music beat. To solve this problem, we further develop a temporal alignment algorithm to align the rhythm of music and dance. Finally, the processed pose sequence is used to synthesize realistic dancing videos in the imagination module. The generated dancing videos match the content and rhythm of the music. Experimental results and subjective evaluations show that the proposed approach can perform the function of generating promising dancing videos by inputting music.

Zeyuan Wang, Yifan Zhao, Jia Li et al.

Given base classes with sufficient labeled samples, the target of few-shot classification is to recognize unlabeled samples of novel classes with only a few labeled samples. Most existing methods only pay attention to the relationship between labeled and unlabeled samples of novel classes, which do not make full use of information within base classes. In this paper, we make two contributions to investigate the few-shot classification problem. First, we report a simple and effective baseline trained on base classes in the way of traditional supervised learning, which can achieve comparable results to the state of the art. Second, based on the baseline, we propose a cooperative bi-path metric for classification, which leverages the correlations between base classes and novel classes to further improve the accuracy. Experiments on two widely used benchmarks show that our method is a simple and effective framework, and a new state of the art is established in the few-shot classification field.

Jiawei Zhao, Yifan Zhao, Jia Li et al.

Salient object detection (SOD) is a crucial and preliminary task for many computer vision applications, which have made progress with deep CNNs. Most of the existing methods mainly rely on the RGB information to distinguish the salient objects, which faces difficulties in some complex scenarios. To solve this, many recent RGBD-based networks are proposed by adopting the depth map as an independent input and fuse the features with RGB information. Taking the advantages of RGB and RGBD methods, we propose a novel depth-aware salient object detection framework, which has following superior designs: 1) It only takes the depth information as training data while only relies on RGB information in the testing phase. 2) It comprehensively optimizes SOD features with multi-level depth-aware regularizations. 3) The depth information also serves as error-weighted map to correct the segmentation process. With these insightful designs combined, we make the first attempt in realizing an unified depth-aware framework with only RGB information as input for inference, which not only surpasses the state-of-the-art performances on five public RGB SOD benchmarks, but also surpasses the RGBD-based methods on five benchmarks by a large margin, while adopting less information and implementation light-weighted. The code and model will be publicly available.

Jia Li, Jinming Su, Changqun Xia et al.

By the aid of attention mechanisms to weight the image features adaptively, recent advanced deep learning-based models encourage the predicted results to approximate the ground-truth masks with as large predictable areas as possible, thus achieving the state-of-the-art performance. However, these methods do not pay enough attention to small areas prone to misprediction. In this way, it is still tough to accurately locate salient objects due to the existence of regions with indistinguishable foreground and background and regions with complex or fine structures. To address these problems, we propose a novel convolutional neural network with purificatory mechanism and structural similarity loss. Specifically, in order to better locate preliminary salient objects, we first introduce the promotion attention, which is based on spatial and channel attention mechanisms to promote attention to salient regions. Subsequently, for the purpose of restoring the indistinguishable regions that can be regarded as error-prone regions of one model, we propose the rectification attention, which is learned from the areas of wrong prediction and guide the network to focus on error-prone regions thus rectifying errors. Through these two attentions, we use the Purificatory Mechanism to impose strict weights with different regions of the whole salient objects and purify results from hard-to-distinguish regions, thus accurately predicting the locations and details of salient objects. In addition to paying different attention to these hard-to-distinguish regions, we also consider the structural constraints on complex regions and propose the Structural Similarity Loss. In experiments, the proposed approach outperforms 19 state-of-the-art methods on six datasets with a notable margin at over 27FPS on a single NVIDIA 1080Ti GPU.

Jia Li, Tong Shen, Wei Zhang et al.

The stunning progress in face manipulation methods has made it possible to synthesize realistic fake face images, which poses potential threats to our society. It is urgent to have face forensics techniques to distinguish those tampered images. A large scale dataset "FaceForensics++" has provided enormous training data generated from prominent face manipulation methods to facilitate anti-fake research. However, previous works focus more on casting it as a classification problem by only considering a global prediction. Through investigation to the problem, we find that training a classification network often fails to capture high quality features, which might lead to sub-optimal solutions. In this paper, we zoom in on the problem by conducting a pixel-level analysis, i.e. formulating it as a pixel-level segmentation task. By evaluating multiple architectures on both segmentation and classification tasks, We show the superiority of viewing the problem from a segmentation perspective. Different ablation studies are also performed to investigate what makes an effective and efficient anti-fake model. Strong baselines are also established, which, we hope, could shed some light on the field of face forensics.

Jia Li, Wen Su, Zengfu Wang

We rethink a well-know bottom-up approach for multi-person pose estimation and propose an improved one. The improved approach surpasses the baseline significantly thanks to (1) an intuitional yet more sensible representation, which we refer to as body parts to encode the connection information between keypoints, (2) an improved stacked hourglass network with attention mechanisms, (3) a novel focal L2 loss which is dedicated to hard keypoint and keypoint association (body part) mining, and (4) a robust greedy keypoint assignment algorithm for grouping the detected keypoints into individual poses. Our approach not only works straightforwardly but also outperforms the baseline by about 15% in average precision and is comparable to the state of the art on the MS-COCO test-dev dataset. The code and pre-trained models are publicly available online.

Limeng Qiao, Yemin Shi, Jia Li et al.

Few-shot learning, which aims at extracting new concepts rapidly from extremely few examples of novel classes, has been featured into the meta-learning paradigm recently. Yet, the key challenge of how to learn a generalizable classifier with the capability of adapting to specific tasks with severely limited data still remains in this domain. To this end, we propose a Transductive Episodic-wise Adaptive Metric (TEAM) framework for few-shot learning, by integrating the meta-learning paradigm with both deep metric learning and transductive inference. With exploring the pairwise constraints and regularization prior within each task, we explicitly formulate the adaptation procedure into a standard semi-definite programming problem. By solving the problem with its closed-form solution on the fly with the setup of transduction, our approach efficiently tailors an episodic-wise metric for each task to adapt all features from a shared task-agnostic embedding space into a more discriminative task-specific metric space. Moreover, we further leverage an attention-based bi-directional similarity strategy for extracting the more robust relationship between queries and prototypes. Extensive experiments on three benchmark datasets show that our framework is superior to other existing approaches and achieves the state-of-the-art performance in the few-shot literature.

Jia Li, Jinming Su, Changqun Xia et al.

Image-based salient object detection (SOD) has been extensively explored in the past decades. However, SOD on 360$^\circ$ omnidirectional images is less studied owing to the lack of datasets with pixel-level annotations. Toward this end, this paper proposes a 360$^\circ$ image-based SOD dataset that contains 500 high-resolution equirectangular images. We collect the representative equirectangular images from five mainstream 360$^\circ$ video datasets and manually annotate all objects and regions over these images with precise masks with a free-viewpoint way. To the best of our knowledge, it is the first public available dataset for salient object detection on 360$^\circ$ scenes. By observing this dataset, we find that distortion from projection, large-scale complex scene and small salient objects are the most prominent characteristics. Inspired by these foundings, this paper proposes a baseline model for SOD on equirectangular images. In the proposed approach, we construct a distortion-adaptive module to deal with the distortion caused by the equirectangular projection. In addition, a multi-scale contextual integration block is introduced to perceive and distinguish the rich scenes and objects in omnidirectional scenes. The whole network is organized in a progressively manner with deep supervision. Experimental results show the proposed baseline approach outperforms the top-performanced state-of-the-art methods on 360$^\circ$ SOD dataset. Moreover, benchmarking results of the proposed baseline approach and other methods on 360$^\circ$ SOD dataset show the proposed dataset is very challenging, which also validate the usefulness of the proposed dataset and approach to boost the development of SOD on 360$^\circ$ omnidirectional scenes.

Yi Zheng, Yifan Zhao, Mengyuan Ren et al.

Recent years have witnessed increasing attention in cartoon media, powered by the strong demands of industrial applications. As the first step to understand this media, cartoon face recognition is a crucial but less-explored task with few datasets proposed. In this work, we first present a new challenging benchmark dataset, consisting of 389,678 images of 5,013 cartoon characters annotated with identity, bounding box, pose, and other auxiliary attributes. The dataset, named iCartoonFace, is currently the largest-scale, high-quality, richannotated, and spanning multiple occurrences in the field of image recognition, including near-duplications, occlusions, and appearance changes. In addition, we provide two types of annotations for cartoon media, i.e., face recognition, and face detection, with the help of a semi-automatic labeling algorithm. To further investigate this challenging dataset, we propose a multi-task domain adaptation approach that jointly utilizes the human and cartoon domain knowledge with three discriminative regularizations. We hence perform a benchmark analysis of the proposed dataset and verify the superiority of the proposed approach in the cartoon face recognition task. We believe this public availability will attract more research attention in broad practical application scenarios.

Lichao Sun, Yingbo Zhou, Ji Wang et al.

Privacy-preserving deep learning is crucial for deploying deep neural network based solutions, especially when the model works on data that contains sensitive information. Most privacy-preserving methods lead to undesirable performance degradation. Ensemble learning is an effective way to improve model performance. In this work, we propose a new method for teacher ensembles that uses more informative network outputs under differential private stochastic gradient descent and provide provable privacy guarantees. Out method employs knowledge distillation and hint learning on intermediate representations to facilitate the training of student model. Additionally, we propose a simple weighted ensemble scheme that works more robustly across different teaching settings. Experimental results on three common image datasets benchmark (i.e., CIFAR10, MINST, and SVHN) demonstrate that our approach outperforms previous state-of-the-art methods on both performance and privacy-budget.

Kaiwen Yu, Jia Li, Yu Zhang et al.

Along with the development of virtual reality (VR), omnidirectional images play an important role in producing multimedia content with immersive experience. However, despite various existing approaches for omnidirectional image stitching, how to quantitatively assess the quality of stitched images is still insufficiently explored. To address this problem, we establish a novel omnidirectional image dataset containing stitched images as well as dual-fisheye images captured from standard quarters of 0$^\circ$, 90$^\circ$, 180$^\circ$ and 270$^\circ$. In this manner, when evaluating the quality of an image stitched from a pair of fisheye images (e.g., 0$^\circ$ and 180$^\circ$), the other pair of fisheye images (e.g., 90$^\circ$ and 270$^\circ$) can be used as the cross-reference to provide ground-truth observations of the stitching regions. Based on this dataset, we further benchmark six widely used stitching models with seven evaluation metrics for IQA. To the best of our knowledge, it is the first dataset that focuses on assessing the stitching quality of omnidirectional images.

Kui Fu, Peipei Shi, Yafei Song et al.

Large convolutional neural network models have recently demonstrated impressive performance on video attention prediction. Conventionally, these models are with intensive computation and large memory. To address these issues, we design an extremely light-weight network with ultrafast speed, named UVA-Net. The network is constructed based on depth-wise convolutions and takes low-resolution images as input. However, this straight-forward acceleration method will decrease performance dramatically. To this end, we propose a coupled knowledge distillation strategy to augment and train the network effectively. With this strategy, the model can further automatically discover and emphasize implicit useful cues contained in the data. Both spatial and temporal knowledge learned by the high-resolution complex teacher networks also can be distilled and transferred into the proposed low-resolution light-weight spatiotemporal network. Experimental results show that the performance of our model is comparable to 11 state-of-the-art models in video attention prediction, while it costs only 0.68 MB memory footprint, runs about 10,106 FPS on GPU and 404 FPS on CPU, which is 206 times faster than previous models.

Shiming Ge, Shengwei Zhao, Chenyu Li et al.

Typically, the deployment of face recognition models in the wild needs to identify low-resolution faces with extremely low computational cost. To address this problem, a feasible solution is compressing a complex face model to achieve higher speed and lower memory at the cost of minimal performance drop. Inspired by that, this paper proposes a learning approach to recognize low-resolution faces via selective knowledge distillation. In this approach, a two-stream convolutional neural network (CNN) is first initialized to recognize high-resolution faces and resolution-degraded faces with a teacher stream and a student stream, respectively. The teacher stream is represented by a complex CNN for high-accuracy recognition, and the student stream is represented by a much simpler CNN for low-complexity recognition. To avoid significant performance drop at the student stream, we then selectively distil the most informative facial features from the teacher stream by solving a sparse graph optimization problem, which are then used to regularize the fine-tuning process of the student stream. In this way, the student stream is actually trained by simultaneously handling two tasks with limited computational resources: approximating the most informative facial cues via feature regression, and recovering the missing facial cues via low-resolution face classification. Experimental results show that the student stream performs impressively in recognizing low-resolution faces and costs only 0.15MB memory and runs at 418 faces per second on CPU and 9,433 faces per second on GPU.

Kui Fu, Jia Li, Yu Zhang et al.

As an emerging vision platform, a drone can look from many abnormal viewpoints which brings many new challenges into the classic vision task of video saliency prediction. To investigate these challenges, this paper proposes a large-scale video dataset for aerial saliency prediction, which consists of ground-truth salient object regions of 1,000 aerial videos, annotated by 24 subjects. To the best of our knowledge, it is the first large-scale video dataset that focuses on visual saliency prediction on drones. Based on this dataset, we propose a Model-guided Multi-path Network (MM-Net) that serves as a baseline model for aerial video saliency prediction. Inspired by the annotation process in eye-tracking experiments, MM-Net adopts multiple information paths, each of which is initialized under the guidance of a classic saliency model. After that, the visual saliency knowledge encoded in the most representative paths is selected and aggregated to improve the capability of MM-Net in predicting spatial saliency in aerial scenarios. Finally, these spatial predictions are adaptively combined with the temporal saliency predictions via a spatiotemporal optimization algorithm. Experimental results show that MM-Net outperforms ten state-of-the-art models in predicting aerial video saliency.

Jia Li, Yafei Song, Jianfeng Zhu et al.

Many advances of deep learning techniques originate from the efforts of addressing the image classification task on large-scale datasets. However, the construction of such clean datasets is costly and time-consuming since the Internet is overwhelmed by noisy images with inadequate and inaccurate tags. In this paper, we propose a Ubiquitous Reweighting Network (URNet) that learns an image classification model from large-scale noisy data. By observing the web data, we find that there are five key challenges, i.e., imbalanced class sizes, high intra-classes diversity and inter-class similarity, imprecise instances, insufficient representative instances, and ambiguous class labels. To alleviate these challenges, we assume that every training instance has the potential to contribute positively by alleviating the data bias and noise via reweighting the influence of each instance according to different class sizes, large instance clusters, its confidence, small instance bags and the labels. In this manner, the influence of bias and noise in the web data can be gradually alleviated, leading to the steadily improving performance of URNet. Experimental results in the WebVision 2018 challenge with 16 million noisy training images from 5000 classes show that our approach outperforms state-of-the-art models and ranks the first place in the image classification task.

Lishi Zhang, Chenghan Fu, Jia Li

Per-pixel masks of semantic objects are very useful in many applications, which, however, are tedious to be annotated. In this paper, we propose a human-agent collaborative annotation approach that can efficiently generate per-pixel masks of semantic objects in tagged images with multi-granularity supervisions. Given a set of tagged image, a computer agent is first dynamically generated to roughly localize the semantic objects described by the tag. The agent first extracts massive object proposals from an image and then infer the tag-related ones under the weak and strong supervisions from linguistically and visually similar images and previously annotated object masks. By representing such supervisions by over-complete dictionaries, the tag-related object proposals can pop-out according to their sparse coding length, which are then converted to superpixels with binary labels. After that, human annotators participate in the annotation process by flipping labels and dividing superpixels with mouse clicks, which are used as click supervisions that teach the agent to recover false positives/negatives in processing images with the same tags. Experimental results show that our approach can facilitate the annotation process and generate object masks that are highly consistent with those generated by the LabelMe toolbox.

Changqun Xia, Jia Li, Jinming Su et al.

In the area of human fixation prediction, dozens of computational saliency models are proposed to reveal certain saliency characteristics under different assumptions and definitions. As a result, saliency model benchmarking often requires several evaluation metrics to simultaneously assess saliency models from multiple perspectives. However, most computational metrics are not designed to directly measure the perceptual similarity of saliency maps so that the evaluation results may be sometimes inconsistent with the subjective impression. To address this problem, this paper first conducts extensive subjective tests to find out how the visual similarities between saliency maps are perceived by humans. Based on the crowdsourced data collected in these tests, we conclude several key factors in assessing saliency maps and quantize the performance of existing metrics. Inspired by these factors, we propose to learn a saliency evaluation metric based on a two-stream convolutional neural network using crowdsourced perceptual judgements. Specifically, the relative saliency score of each pair from the crowdsourced data is utilized to regularize the network during the training process. By capturing the key factors shared by various subjects in comparing saliency maps, the learned metric better aligns with human perception of saliency maps, making it a good complement to the existing metrics. Experimental results validate that the learned metric can be generalized to the comparisons of saliency maps from new images, new datasets, new models and synthetic data. Due to the effectiveness of the learned metric, it also can be used to facilitate the development of new models for fixation prediction.

Wei Teng, Yu Zhang, Xiaowu Chen et al.

Image co-segmentation is a challenging task in computer vision that aims to segment all pixels of the objects from a predefined semantic category. In real-world cases, however, common foreground objects often vary greatly in appearance, making their global shapes highly inconsistent across images and difficult to be segmented. To address this problem, this paper proposes a novel co-segmentation approach that transfers patch-level local object shapes which appear more consistent across different images. In our framework, a multi-scale patch neighbourhood system is first generated using proposal flow on arbitrary image-pair, which is further refined by Locally Linear Embedding. Based on the patch relationships, we propose an efficient algorithm to jointly segment the objects in each image while transferring their local shapes across different images. Extensive experiments demonstrate that the proposed method can robustly and effectively segment common objects from an image set. On iCoseg, MSRC and Coseg-Rep dataset, the proposed approach performs comparable or better than the state-of-thearts, while on a more challenging benchmark Fashionista dataset, our method achieves significant improvements.

Shenlong Wang, Linjie Luo, Ning Zhang et al.

Finding visual correspondence between local features is key to many computer vision problems. While defining features with larger contextual scales usually implies greater discriminativeness, it could also lead to less spatial accuracy of the features. We propose AutoScaler, a scale-attention network to explicitly optimize this trade-off in visual correspondence tasks. Our network consists of a weight-sharing feature network to compute multi-scale feature maps and an attention network to combine them optimally in the scale space. This allows our network to have adaptive receptive field sizes over different scales of the input. The entire network is trained end-to-end in a siamese framework for visual correspondence tasks. Our method achieves favorable results compared to state-of-the-art methods on challenging optical flow and semantic matching benchmarks, including Sintel, KITTI and CUB-2011. We also show that our method can generalize to improve hand-crafted descriptors (e.g Daisy) on general visual correspondence tasks. Finally, our attention network can generate visually interpretable scale attention maps.