Hui Yu

Jiaxin Wang, Muwei Jian, Hui Yu et al.

Facial Expression Recognition (FER) in the wild is still challenging due to uncontrolled variations in pose, occlusion, and illumination. Most existing attention-based methods primarily rely on visual appearance cues, suffering from attention redundancy and instability, which limits their performance in complex scenarios. To address these issues, we propose a novel landmark-guided contrastive learning network with vision-language enhancement for FER (LaCoVL-FER), which integrates geometric priors from facial landmarks and semantic priors from a vision-language model. Specifically, a Landmark-Guided Adaptive Encoder (LGAE) is designed to introduce geometric priors through a Bi-branch Gated Cross Attention (BGCA) mechanism, which achieves adaptive fusion of landmark-based geometric and visual appearance features to produce expression-relevant features, thereby focusing on key facial regions and suppressing noise interference. In parallel, a Vision-Language Enhancement Strategy (VLES) is presented to leverage the expression-relevant features to refine the generalizable visual features extracted by the frozen pretrained CLIP image encoder, yielding expression-specific visual representations. Based on these representations, an Expression-Conditioned Prompting (ECP) mechanism is utilized to further adapt the textual features of fixed class-level prompts from the frozen pretrained CLIP text encoder, generating more instance-aware textual representations. These visual-textual representations are aligned as semantic priors to enhance the robustness and generalization of FER. Quantitative and qualitative experiments demonstrate that our LaCoVL-FER outperforms state-of-the-art methods on three representative real-world FER datasets, including RAF-DB, FERPlus, and AffectNet. The code is available at https://github.com/ylin06804/LaCoVL-FER.

Wei Zhang, Yueyue Jiang, Junyu Dong et al.

Forecasts by the European Centre for Medium-Range Weather Forecasts (ECMWF; EC for short) can provide a basis for the establishment of maritime-disaster warning systems, but they contain some systematic biases.The fifth-generation EC atmospheric reanalysis (ERA5) data have high accuracy, but are delayed by about 5 days. To overcome this issue, a spatiotemporal deep-learning method could be used for nonlinear mapping between EC and ERA5 data, which would improve the quality of EC wind forecast data in real time. In this study, we developed the Multi-Task-Double Encoder Trajectory Gated Recurrent Unit (MT-DETrajGRU) model, which uses an improved double-encoder forecaster architecture to model the spatiotemporal sequence of the U and V components of the wind field; we designed a multi-task learning loss function to correct wind speed and wind direction simultaneously using only one model. The study area was the western North Pacific (WNP), and real-time rolling bias corrections were made for 10-day wind-field forecasts released by the EC between December 2020 and November 2021, divided into four seasons. Compared with the original EC forecasts, after correction using the MT-DETrajGRU model the wind speed and wind direction biases in the four seasons were reduced by 8-11% and 9-14%, respectively. In addition, the proposed method modelled the data uniformly under different weather conditions. The correction performance under normal and typhoon conditions was comparable, indicating that the data-driven mode constructed here is robust and generalizable.

Zixiao Wang, Yuluo Guo, Jin Zhao et al.

In this paper, we propose a Graph Inception Diffusion Networks(GIDN) model. This model generalizes graph diffusion in different feature spaces, and uses the inception module to avoid the large amount of computations caused by complex network structures. We evaluate GIDN model on Open Graph Benchmark(OGB) datasets, reached an 11% higher performance than AGDN on ogbl-collab dataset.

Pierre Degond, Silke Henkes, Hui Yu

Motivated by recent experimental and computational results that show a motility-induced clustering transition in self-propelled particle systems, we study an individual model and its corresponding Self-Organized Hydrodynamic model for collective behaviour that incorporates a density-dependent velocity, as well as inter-particle alignment. The modal analysis of the hydrodynamic model elucidates the relationship between the stability of the equilibria and the changing velocity, and the formation of clusters. We find, in agreement with earlier results for non-aligning particles, that the key criterion for stability is $(ρv(ρ))'> 0$, i.e. a non-rapid decrease of velocity with density. Numerical simulation for both the individual and hydrodynamic models with a velocity function inspired by experiment demonstrates the validity of the theoretical results.

Kaifang Long, Jikun Dong, Shengyu Fan et al.

Recently, with the continuous development of deep learning, the performance of named entity recognition tasks has been dramatically improved. However, the privacy and the confidentiality of data in some specific fields, such as biomedical and military, cause insufficient data to support the training of deep neural networks. In this paper, we propose an encryption learning framework to address the problems of data leakage and inconvenient disclosure of sensitive data in certain domains. We introduce multiple encryption algorithms to encrypt training data in the named entity recognition task for the first time. In other words, we train the deep neural network using the encrypted data. We conduct experiments on six Chinese datasets, three of which are constructed by ourselves. The experimental results show that the encryption method achieves satisfactory results. The performance of some models trained with encrypted data even exceeds the performance of the unencrypted method, which verifies the effectiveness of the introduced encryption method and solves the problem of data leakage to a certain extent.

Ziyuan Yang, Zerui Shao, Huijie Huangfu et al.

Federated learning (FL) is a promising distributed paradigm, eliminating the need for data sharing but facing challenges from data heterogeneity. Personalized parameter generation through a hypernetwork proves effective, yet existing methods fail to personalize local model structures. This leads to redundant parameters struggling to adapt to diverse data distributions. To address these limitations, we propose FedOFA, utilizing personalized orthogonal filter attention for parameter recalibration. The core is the Two-stream Filter-aware Attention (TFA) module, meticulously designed to extract personalized filter-aware attention maps, incorporating Intra-Filter Attention (IntraFa) and Inter-Filter Attention (InterFA) streams. These streams enhance representation capability and explore optimal implicit structures for local models. Orthogonal regularization minimizes redundancy by averting inter-correlation between filters. Furthermore, we introduce an Attention-Guided Pruning Strategy (AGPS) for communication efficiency. AGPS selectively retains crucial neurons while masking redundant ones, reducing communication costs without performance sacrifice. Importantly, FedOFA operates on the server side, incurring no additional computational cost on the client, making it advantageous in communication-constrained scenarios. Extensive experiments validate superior performance over state-of-the-art approaches, with code availability upon paper acceptance.

Qingxuan Lv, Junyu Dong, Yuezun Li et al.

Despite recent advances in stereo matching, the extension to intricate underwater settings remains unexplored, primarily owing to: 1) the reduced visibility, low contrast, and other adverse effects of underwater images; 2) the difficulty in obtaining ground truth data for training deep learning models, i.e. simultaneously capturing an image and estimating its corresponding pixel-wise depth information in underwater environments. To enable further advance in underwater stereo matching, we introduce a large synthetic dataset called UWStereo. Our dataset includes 29,568 synthetic stereo image pairs with dense and accurate disparity annotations for left view. We design four distinct underwater scenes filled with diverse objects such as corals, ships and robots. We also induce additional variations in camera model, lighting, and environmental effects. In comparison with existing underwater datasets, UWStereo is superior in terms of scale, variation, annotation, and photo-realistic image quality. To substantiate the efficacy of the UWStereo dataset, we undertake a comprehensive evaluation compared with nine state-of-the-art algorithms as benchmarks. The results indicate that current models still struggle to generalize to new domains. Hence, we design a new strategy that learns to reconstruct cross domain masked images before stereo matching training and integrate a cross view attention enhancement module that aggregates long-range content information to enhance the generalization ability.

Jianzhang Zheng, Hao Shen, Jian Yang et al.

Autoencoders have achieved great success in various computer vision applications. The autoencoder learns appropriate low dimensional image representations through the self-supervised paradigm, i.e., reconstruction. Existing studies mainly focus on the minimizing the reconstruction error on pixel level of image, while ignoring the preservation of Intrinsic Dimension (ID), which is a fundamental geometric property of data representations in Deep Neural Networks (DNNs). Motivated by the important role of ID, in this paper, we propose a novel deep representation learning approach with autoencoder, which incorporates regularization of the global and local ID constraints into the reconstruction of data representations. This approach not only preserves the global manifold structure of the whole dataset, but also maintains the local manifold structure of the feature maps of each point, which makes the learned low-dimensional features more discriminant and improves the performance of the downstream algorithms. To our best knowledge, existing works are rare and limited on exploiting both global and local ID invariant properties on the regularization of autoencoders. Numerical experimental results on benchmark datasets (Extended Yale B, Caltech101 and ImageNet) show that the resulting regularized learning models achieve better discriminative representations for downstream tasks including image classification and clustering.

Wei Wu, Fan Lu, Yunnan Wang et al.

Offering great potential in robotic manipulation, a capable Vision-Language-Action (VLA) foundation model is expected to faithfully generalize across tasks and platforms while ensuring cost efficiency (e.g., data and GPU hours required for adaptation). To this end, we develop LingBot-VLA with around 20,000 hours of real-world data from 9 popular dual-arm robot configurations. Through a systematic assessment on 3 robotic platforms, each completing 100 tasks with 130 post-training episodes per task, our model achieves clear superiority over competitors, showcasing its strong performance and broad generalizability. We have also built an efficient codebase, which delivers a throughput of 261 samples per second per GPU with an 8-GPU training setup, representing a 1.5~2.8$\times$ (depending on the relied VLM base model) speedup over existing VLA-oriented codebases. The above features ensure that our model is well-suited for real-world deployment. To advance the field of robot learning, we provide open access to the code, base model, and benchmark data, with a focus on enabling more challenging tasks and promoting sound evaluation standards.

Zhe Chang, Haodong Jin, Ying Sun et al.

High-fidelity 4D dynamic facial avatar reconstruction from monocular video is a critical yet challenging task, driven by increasing demands for immersive virtual human applications. While Neural Radiance Fields (NeRF) have advanced scene representation, their capacity to capture high-frequency facial details, such as dynamic wrinkles and subtle textures from information-constrained monocular streams, requires significant enhancement. To tackle this challenge, we propose a novel hybrid neural radiance field framework, called Geometry-Aware-Transformer Enhanced NeRF (GAT-NeRF) for high-fidelity and controllable 4D facial avatar reconstruction, which integrates the Transformer mechanism into the NeRF pipeline. GAT-NeRF synergistically combines a coordinate-aligned Multilayer Perceptron (MLP) with a lightweight Transformer module, termed as Geometry-Aware-Transformer (GAT) due to its processing of multi-modal inputs containing explicit geometric priors. The GAT module is enabled by fusing multi-modal input features, including 3D spatial coordinates, 3D Morphable Model (3DMM) expression parameters, and learnable latent codes to effectively learn and enhance feature representations pertinent to fine-grained geometry. The Transformer's effective feature learning capabilities are leveraged to significantly augment the modeling of complex local facial patterns like dynamic wrinkles and acne scars. Comprehensive experiments unequivocally demonstrate GAT-NeRF's state-of-the-art performance in visual fidelity and high-frequency detail recovery, forging new pathways for creating realistic dynamic digital humans for multimedia applications.

Mehran Taghian, Yunke Peng, Xing Huang et al.

Large foundation models have become central to modern machine learning, with performance scaling predictably with model size and data. However, training and deploying such models incur substantial computational and memory costs, motivating the development of low-precision training techniques. Recent work has demonstrated that 4-bit floating-point (FP4) formats--such as MXFP4 and NVFP4--can be successfully applied to linear GEMM operations in large language models (LLMs), achieving up to 4x improvements in compute throughput and memory efficiency compared to higher-precision baselines. In this work, we investigate the recently proposed HiFloat4 FP4 format for Huawei Ascend NPUs and systematically compare it with MXFP4 in large-scale training settings. All experiments are conducted on Ascend NPU clusters, with linear and expert GEMM operations performed entirely in FP4 precision. We evaluate both dense architectures (e.g., Pangu and LLaMA-style models) and mixture-of-experts (MoE) models, where both standard linear layers and expert-specific GEMMs operate in FP4. Furthermore, we explore stabilization techniques tailored to FP4 training that significantly reduce numerical degradation, maintaining relative error within 1% of full-precision baselines while preserving the efficiency benefits of 4-bit computation. Our results provide a comprehensive empirical study of FP4 training on NPUs and highlight the practical trade-offs between FP4 formats in large-scale dense and MoE models.

Yuqiao Zeng, Xu Wang, Tengfei Liang et al.

Multimodal learning integrates complementary information from different modalities such as image, text, and audio to improve model performance, but its success relies on large-scale labeled data, which is costly to obtain. Active learning (AL) mitigates this challenge by selectively annotating informative samples. In multimodal settings, many approaches implicitly assume that modality importance is stable across rounds and keep selection rules fixed at the fusion stage, which leaves them insensitive to the dynamic nature of multimodal learning, where the relative value of modalities and the difficulty of instances shift as training proceeds. To address this issue, we propose RL-MBA, a reinforcement-learning framework for modality-balanced, difficulty-aware multimodal active learning. RL-MBA models sample selection as a Markov Decision Process, where the policy adapts to modality contributions, uncertainty, and diversity, and the reward encourages accuracy gains and balance. Two key components drive this adaptability: (1) Adaptive Modality Contribution Balancing (AMCB), which dynamically adjusts modality weights via reinforcement feedback, and (2) Evidential Fusion for DifficultyAware Policy Adjustment (EFDA), which estimates sample difficulty via uncertainty-based evidential fusion to prioritize informative samples. Experiments on Food101, KineticsSound, and VGGSound demonstrate that RL-MBA consistently outperforms strong baselines, improving both classification accuracy and modality fairness under limited labeling budgets.

Ziyuan Yang, Huijie Huangfu, Maosong Ran et al.

Deep learning (DL) has made significant advancements in tomographic imaging, particularly in low-dose computed tomography (LDCT) denoising. A recent trend involves servers training powerful models with large amounts of self-collected private data and providing application programming interfaces (APIs) for users, such as Chat-GPT. To avoid model leakage, users are required to upload their data to the server model, but this way raises public concerns about the potential risk of privacy disclosure, especially for medical data. Hence, to alleviate related concerns, in this paper, we propose to directly denoise LDCT in the encrypted domain to achieve privacy-preserving cloud services without exposing private data to the server. To this end, we employ homomorphic encryption to encrypt private LDCT data, which is then transferred to the server model trained with plaintext LDCT for further denoising. However, since traditional operations, such as convolution and linear transformation, in DL methods cannot be directly used in the encrypted domain, we transform the fundamental mathematic operations in the plaintext domain into the operations in the encrypted domain. In addition, we present two interactive frameworks for linear and nonlinear models in this paper, both of which can achieve lossless operating. In this way, the proposed methods can achieve two merits, the data privacy is well protected and the server model is free from the risk of model leakage. Moreover, we provide theoretical proof to validate the lossless property of our framework. Finally, experiments were conducted to demonstrate that the transferred contents are well protected and cannot be reconstructed. The code will be released once the paper is accepted.

Yili Chen, Zhengyu Li, Zheng Wan et al.

AI-driven drug design relies significantly on predicting molecular properties, which is a complex task. In current approaches, the most commonly used feature representations for training deep neural network models are based on SMILES and molecular graphs. While these methods are concise and efficient, they have limitations in capturing complex spatial information. Recently, researchers have recognized the importance of incorporating three-dimensional information of molecular structures into models. However, capturing spatial information requires the introduction of additional units in the generator, bringing additional design and computational costs. Therefore, it is necessary to develop a method for predicting molecular properties that effectively combines spatial structural information while maintaining the simplicity and efficiency of graph neural networks. In this work, we propose an embedding approach CTAGE, utilizing $k$-hop discrete Ricci curvature to extract structural insights from molecular graph data. This effectively integrates spatial structural information while preserving the training complexity of the network. Experimental results indicate that introducing node curvature significantly improves the performance of current graph neural network frameworks, validating that the information from k-hop node curvature effectively reflects the relationship between molecular structure and function.

Haizhou Liu, Haodong Jin, Yiming Wang et al.

Video summarization is a crucial technique for social understanding, enabling efficient browsing of massive multimedia content and extraction of key information from social platforms. Most existing unsupervised summarization methods rely on Generative Adversarial Networks (GANs) to enhance keyframe selection and generate coherent, video summaries through adversarial training. However, such approaches primarily exploit unimodal features, overlooking the guiding role of semantic information in keyframe selection, and often suffer from unstable training. To address these limitations, we propose a novel Semantic-Guided Unsupervised Video Summarization method. Specifically, we design a novel frame-level semantic alignment attention mechanism and integrate it into a keyframe selector, which guides the Transformer-based generator within the adversarial framework to better reconstruct videos. In addition, we adopt an incremental training strategy to progressively update the model components, effectively mitigating the instability of GAN training. Experimental results demonstrate that our approach achieves superior performance on multiple benchmark datasets.

Chao Song, Zhiyuan Liu, Han Huang et al.

Designing enzyme backbones with substrate-specific functionality is a critical challenge in computational protein engineering. Current generative models excel in protein design but face limitations in binding data, substrate-specific control, and flexibility for de novo enzyme backbone generation. To address this, we introduce EnzyBind, a dataset with 11,100 experimentally validated enzyme-substrate pairs specifically curated from PDBbind. Building on this, we propose EnzyControl, a method that enables functional and substrate-specific control in enzyme backbone generation. Our approach generates enzyme backbones conditioned on MSA-annotated catalytic sites and their corresponding substrates, which are automatically extracted from curated enzyme-substrate data. At the core of EnzyControl is EnzyAdapter, a lightweight, modular component integrated into a pretrained motif-scaffolding model, allowing it to become substrate-aware. A two-stage training paradigm further refines the model's ability to generate accurate and functional enzyme structures. Experiments show that our EnzyControl achieves the best performance across structural and functional metrics on EnzyBind and EnzyBench benchmarks, with particularly notable improvements of 13\% in designability and 13\% in catalytic efficiency compared to the baseline models. The code is released at https://github.com/Vecteur-libre/EnzyControl.

Yingyu Chen, Ziyuan Yang, Ming Yan et al.

Multi-modality (MM) semi-supervised learning (SSL) based medical image segmentation has recently gained increasing attention for its ability to utilize MM data and reduce reliance on labeled images. However, current methods face several challenges: (1) Complex network designs hinder scalability to scenarios with more than two modalities. (2) Focusing solely on modality-invariant representation while neglecting modality-specific features, leads to incomplete MM learning. (3) Leveraging unlabeled data with generative methods can be unreliable for SSL. To address these problems, we propose Double Bank Dual Consistency (DBDC), a novel MM-SSL approach for medical image segmentation. To address challenge (1), we propose a modality all-in-one segmentation network that accommodates data from any number of modalities, removing the limitation on modality count. To address challenge (2), we design two learnable plug-in banks, Modality-Level Modulation bank (MLMB) and Modality-Level Prototype (MLPB) bank, to capture both modality-invariant and modality-specific knowledge. These banks are updated using our proposed Modality Prototype Contrastive Learning (MPCL). Additionally, we design Modality Adaptive Weighting (MAW) to dynamically adjust learning weights for each modality, ensuring balanced MM learning as different modalities learn at different rates. Finally, to address challenge (3), we introduce a Dual Consistency (DC) strategy that enforces consistency at both the image and feature levels without relying on generative methods. We evaluate our method on a 2-to-4 modality segmentation task using three open-source datasets, and extensive experiments show that our method outperforms state-of-the-art approaches.

Kun Zhou, Hui Yu, Wayne Xin Zhao et al.

Recently, deep neural networks such as RNN, CNN and Transformer have been applied in the task of sequential recommendation, which aims to capture the dynamic preference characteristics from logged user behavior data for accurate recommendation. However, in online platforms, logged user behavior data is inevitable to contain noise, and deep recommendation models are easy to overfit on these logged data. To tackle this problem, we borrow the idea of filtering algorithms from signal processing that attenuates the noise in the frequency domain. In our empirical experiments, we find that filtering algorithms can substantially improve representative sequential recommendation models, and integrating simple filtering algorithms (eg Band-Stop Filter) with an all-MLP architecture can even outperform competitive Transformer-based models. Motivated by it, we propose \textbf{FMLP-Rec}, an all-MLP model with learnable filters for sequential recommendation task. The all-MLP architecture endows our model with lower time complexity, and the learnable filters can adaptively attenuate the noise information in the frequency domain. Extensive experiments conducted on eight real-world datasets demonstrate the superiority of our proposed method over competitive RNN, CNN, GNN and Transformer-based methods. Our code and data are publicly available at the link: \textcolor{blue}{\url{https://github.com/RUCAIBox/FMLP-Rec}}.

Chenglizhao Chen, Boze Li, Mengke Song et al.

Depression is a severe global mental health issue that impairs daily functioning and overall quality of life. Although recent audio-visual approaches have improved automatic depression detection, methods that ignore emotional cues often fail to capture subtle depressive signals hidden within emotional expressions. Conversely, those incorporating emotions frequently confuse transient emotional expressions with stable depressive symptoms in feature representations, a phenomenon termed \emph{Emotional Ambiguity}, thereby leading to detection errors. To address this critical issue, we propose READ-Net, the first audio-visual depression detection framework explicitly designed to resolve Emotional Ambiguity through Adaptive Feature Recalibration (AFR). The core insight of AFR is to dynamically adjust the weights of emotional features to enhance depression-related signals. Rather than merely overlooking or naively combining emotional information, READ-Net innovatively identifies and preserves depressive-relevant cues within emotional features, while adaptively filtering out irrelevant emotional noise. This recalibration strategy significantly clarifies feature representations, and effectively mitigates the persistent challenge of emotional interference. Additionally, READ-Net can be easily integrated into existing frameworks for improved performance. Extensive evaluations on three publicly available datasets show that READ-Net outperforms state-of-the-art methods, with average gains of 4.55\% in accuracy and 1.26\% in F1-score, demonstrating its robustness to emotional disturbances and improving audio-visual depression detection.

Zhe Chang, Haodong Jin, Yan Song et al.

Creating high-fidelity, real-time drivable 3D head avatars is a core challenge in digital animation. While 3D Gaussian Splashing (3D-GS) offers unprecedented rendering speed and quality, current animation techniques often rely on a "one-size-fits-all" global tuning approach, where all Gaussian primitives are uniformly driven by a single expression code. This simplistic approach fails to unravel the distinct dynamics of different facial regions, such as deformable skin versus rigid teeth, leading to significant blurring and distortion artifacts. We introduce Conditionally-Adaptive Gaussian Avatars (CAG-Avatar), a framework that resolves this key limitation. At its core is a Conditionally Adaptive Fusion Module built on cross-attention. This mechanism empowers each 3D Gaussian to act as a query, adaptively extracting relevant driving signals from the global expression code based on its canonical position. This "tailor-made" conditioning strategy drastically enhances the modeling of fine-grained, localized dynamics. Our experiments confirm a significant improvement in reconstruction fidelity, particularly for challenging regions such as teeth, while preserving real-time rendering performance.

Qingxuan Lv, Yuezun Li, Junyu Dong et al.

Recent DeepFake detection methods have shown excellent performance on public datasets but are significantly degraded on new forgeries. Solving this problem is important, as new forgeries emerge daily with the continuously evolving generative techniques. Many efforts have been made for this issue by seeking the commonly existing traces empirically on data level. In this paper, we rethink this problem and propose a new solution from the unsupervised domain adaptation perspective. Our solution, called DomainForensics, aims to transfer the forgery knowledge from known forgeries to new forgeries. Unlike recent efforts, our solution does not focus on data view but on learning strategies of DeepFake detectors to capture the knowledge of new forgeries through the alignment of domain discrepancies. In particular, unlike the general domain adaptation methods which consider the knowledge transfer in the semantic class category, thus having limited application, our approach captures the subtle forgery traces. We describe a new bi-directional adaptation strategy dedicated to capturing the forgery knowledge across domains. Specifically, our strategy considers both forward and backward adaptation, to transfer the forgery knowledge from the source domain to the target domain in forward adaptation and then reverse the adaptation from the target domain to the source domain in backward adaptation. In forward adaptation, we perform supervised training for the DeepFake detector in the source domain and jointly employ adversarial feature adaptation to transfer the ability to detect manipulated faces from known forgeries to new forgeries. In backward adaptation, we further improve the knowledge transfer by coupling adversarial adaptation with self-distillation on new forgeries. This enables the detector to expose new forgery features from unlabeled data and avoid forgetting the known knowledge of known...

Xiaoqiang Yan, Yingtao Gan, Yiqiao Mao et al.

Multi-view action clustering leverages the complementary information from different camera views to enhance the clustering performance. Although existing approaches have achieved significant progress, they assume all camera views are available in advance, which is impractical when the camera view is incremental over time. Besides, learning the invariant information among multiple camera views is still a challenging issue, especially in continual learning scenario. Aiming at these problems, we propose a novel continual action clustering (CAC) method, which is capable of learning action categories in a continual learning manner. To be specific, we first devise a category memory library, which captures and stores the learned categories from historical views. Then, as a new camera view arrives, we only need to maintain a consensus partition matrix, which can be updated by leveraging the incoming new camera view rather than keeping all of them. Finally, a three-step alternate optimization is proposed, in which the category memory library and consensus partition matrix are optimized. The empirical experimental results on 6 realistic multi-view action collections demonstrate the excellent clustering performance and time/space efficiency of the CAC compared with 15 state-of-the-art baselines.

Adarsh Bhandary Panambur, Hui Yu, Sheethal Bhat et al.

The assessment of breast density is crucial in the context of breast cancer screening, especially in populations with a higher percentage of dense breast tissues. This study introduces a novel data augmentation technique termed Attention-Guided Erasing (AGE), devised to enhance the downstream classification of four distinct breast density categories in mammography following the BI-RADS recommendation in the Vietnamese cohort. The proposed method integrates supplementary information during transfer learning, utilizing visual attention maps derived from a vision transformer backbone trained using the self-supervised DINO method. These maps are utilized to erase background regions in the mammogram images, unveiling only the potential areas of dense breast tissues to the network. Through the incorporation of AGE during transfer learning with varying random probabilities, we consistently surpass classification performance compared to scenarios without AGE and the traditional random erasing transformation. We validate our methodology using the publicly available VinDr-Mammo dataset. Specifically, we attain a mean F1-score of 0.5910, outperforming values of 0.5594 and 0.5691 corresponding to scenarios without AGE and with random erasing (RE), respectively. This superiority is further substantiated by t-tests, revealing a p-value of p<0.0001, underscoring the statistical significance of our approach.

Wei Liang, Hui Yu, Derui Ding et al.

Head avatar reenactment focuses on creating animatable personal avatars from monocular videos, serving as a foundational element for applications like social signal understanding, gaming, human-machine interaction, and computer vision. Recent advances in 3D Morphable Model (3DMM)-based facial reconstruction methods have achieved remarkable high-fidelity face estimation. However, on the one hand, they struggle to capture the entire head, including non-facial regions and background details in real time, which is an essential aspect for producing realistic, high-fidelity head avatars. On the other hand, recent approaches leveraging generative adversarial networks (GANs) for head avatar generation from videos can achieve high-quality reenactments but encounter limitations in reproducing fine-grained head details, such as wrinkles and hair textures. In addition, existing methods generally rely on a large amount of training data, and rarely focus on using only a simple selfie video to achieve avatar reenactment. To address these challenges, this study introduces a method for detailed head avatar reenactment using a selfie video. The approach combines 3DMMs with a StyleGAN-based generator. A detailed reconstruction model is proposed, incorporating mixed loss functions for foreground reconstruction and avatar image generation during adversarial training to recover high-frequency details. Qualitative and quantitative evaluations on self-reenactment and cross-reenactment tasks demonstrate that the proposed method achieves superior head avatar reconstruction with rich and intricate textures compared to existing approaches.

Yuhui Zhang, Hui Yu, Wei Liang et al.

Dynamic Neural Radiance Fields (NeRF) have demonstrated considerable success in generating high-fidelity 3D models of talking portraits. Despite significant advancements in the rendering speed and generation quality, challenges persist in accurately and efficiently capturing mouth movements in talking portraits. To tackle this challenge, we propose an automatic method based on blink embedding and hash grid landmarks encoding in this study, which can substantially enhance the fidelity of talking faces. Specifically, we leverage facial features encoded as conditional features and integrate audio features as residual terms into our model through a Dynamic Landmark Transformer. Furthermore, we employ neural radiance fields to model the entire face, resulting in a lifelike face representation. Experimental evaluations have validated the superiority of our approach to existing methods.

Mengyu Sun, Ziyuan Yang, Yongqiang Huang et al.

Artificial intelligence (AI) has demonstrated considerable potential in the realm of medical imaging. However, the development of high-performance AI models typically necessitates training on large-scale, centralized datasets. This approach is confronted with significant challenges due to strict patient privacy regulations and legal restrictions on data sharing and utilization. These limitations hinder the development of large-scale models in medical domains and impede continuous updates and training with new data. Federated Learning (FL), a privacy-preserving distributed training framework, offers a new solution by enabling collaborative model development across fragmented medical datasets. In this survey, we review FL's contributions at two stages of the full-stack medical analysis pipeline. First, in upstream tasks such as CT or MRI reconstruction, FL enables joint training of robust reconstruction networks on diverse, multi-institutional datasets, alleviating data scarcity while preserving confidentiality. Second, in downstream clinical tasks like tumor diagnosis and segmentation, FL supports continuous model updating by allowing local fine-tuning on new data without centralizing sensitive images. We comprehensively analyze FL implementations across the medical imaging pipeline, from physics-informed reconstruction networks to diagnostic AI systems, highlighting innovations that improve communication efficiency, align heterogeneous data, and ensure secure parameter aggregation. Meanwhile, this paper provides an outlook on future research directions, aiming to serve as a valuable reference for the field's development.

Jinkun Jiang, Qingxuan Lv, Yuezun Li et al.

Source-free Unsupervised Domain Adaptation (SFDA) aims to classify target samples by only accessing a pre-trained source model and unlabelled target samples. Since no source data is available, transferring the knowledge from the source domain to the target domain is challenging. Existing methods normally exploit the pair-wise relation among target samples and attempt to discover their correlations by clustering these samples based on semantic features. The drawback of these methods includes: 1) the pair-wise relation is limited to exposing the underlying correlations of two more samples, hindering the exploration of the structural information embedded in the target domain; 2) the clustering process only relies on the semantic feature, while overlooking the critical effect of domain shift, i.e., the distribution differences between the source and target domains. To address these issues, we propose a new SFDA method that exploits the high-order neighborhood relation and explicitly takes the domain shift effect into account. Specifically, we formulate the SFDA as a Hypergraph learning problem and construct hyperedges to explore the local group and context information among multiple samples. Moreover, we integrate a self-loop strategy into the constructed hypergraph to elegantly introduce the domain uncertainty of each sample. By clustering these samples based on hyperedges, both the semantic feature and domain shift effects are considered. We then describe an adaptive relation-based objective to tune the model with soft attention levels for all samples. Extensive experiments are conducted on Office-31, Office-Home, VisDA, and PointDA-10 datasets. The results demonstrate the superiority of our method over state-of-the-art counterparts.

Hui Yu, ShiYu Zhao, JianYu Shi

Motivation: Computational prediction of multiple-type drug-drug interaction (DDI) helps reduce unexpected side effects in poly-drug treatments. Although existing computational approaches achieve inspiring results, they ignore that the action of a drug is mainly caused by its chemical substructures. In addition, their interpretability is still weak. Results: In this paper, by supposing that the interactions between two given drugs are caused by their local chemical structures (sub-structures) and their DDI types are determined by the linkages between different substructure sets, we design a novel Substructure-ware Tensor Neural Network model for DDI prediction (STNN-DDI). The proposed model learns a 3-D tensor of (substructure, in-teraction type, substructure) triplets, which characterizes a substructure-substructure interaction (SSI) space. According to a list of predefined substructures with specific chemical meanings, the mapping of drugs into this SSI space enables STNN-DDI to perform the multiple-type DDI prediction in both transductive and inductive scenarios in a unified form with an explicable manner. The compar-ison with deep learning-based state-of-the-art baselines demonstrates the superiority of STNN-DDI with the significant improvement of AUC, AUPR, Accuracy, and Precision. More importantly, case studies illustrate its interpretability by both revealing a crucial sub-structure pair across drugs regarding a DDI type of interest and uncovering interaction type-specific substructure pairs in a given DDI. In summary, STNN-DDI provides an effective approach to predicting DDIs as well as explaining the interaction mechanisms among drugs.

Haoran Zhao, Xin Sun, Junyu Dong et al.

Knowledge distillation has been widely used to produce portable and efficient neural networks which can be well applied on edge devices for computer vision tasks. However, almost all top-performing knowledge distillation methods need to access the original training data, which usually has a huge size and is often unavailable. To tackle this problem, we propose a novel data-free approach in this paper, named Dual Discriminator Adversarial Distillation (DDAD) to distill a neural network without any training data or meta-data. To be specific, we use a generator to create samples through dual discriminator adversarial distillation, which mimics the original training data. The generator not only uses the pre-trained teacher's intrinsic statistics in existing batch normalization layers but also obtains the maximum discrepancy from the student model. Then the generated samples are used to train the compact student network under the supervision of the teacher. The proposed method obtains an efficient student network which closely approximates its teacher network, despite using no original training data. Extensive experiments are conducted to to demonstrate the effectiveness of the proposed approach on CIFAR-10, CIFAR-100 and Caltech101 datasets for classification tasks. Moreover, we extend our method to semantic segmentation tasks on several public datasets such as CamVid and NYUv2. All experiments show that our method outperforms all baselines for data-free knowledge distillation.

Haoran Zhao, Kun Gong, Xin Sun et al.

Knowledge distillation is a popular paradigm for learning portable neural networks by transferring the knowledge from a large model into a smaller one. Most existing approaches enhance the student model by utilizing the similarity information between the categories of instance level provided by the teacher model. However, these works ignore the similarity correlation between different instances that plays an important role in confidence prediction. To tackle this issue, we propose a novel method in this paper, called similarity transfer for knowledge distillation (STKD), which aims to fully utilize the similarities between categories of multiple samples. Furthermore, we propose to better capture the similarity correlation between different instances by the mixup technique, which creates virtual samples by a weighted linear interpolation. Note that, our distillation loss can fully utilize the incorrect classes similarities by the mixed labels. The proposed approach promotes the performance of student model as the virtual sample created by multiple images produces a similar probability distribution in the teacher and student networks. Experiments and ablation studies on several public classification datasets including CIFAR-10,CIFAR-100,CINIC-10 and Tiny-ImageNet verify that this light-weight method can effectively boost the performance of the compact student model. It shows that STKD substantially has outperformed the vanilla knowledge distillation and has achieved superior accuracy over the state-of-the-art knowledge distillation methods.

Xiaoxu Cai, Hui Yu, Jianwen Lou et al.

We present to recover the complete 3D facial geometry from a single depth view by proposing an Attention Guided Generative Adversarial Networks (AGGAN). In contrast to existing work which normally requires two or more depth views to recover a full 3D facial geometry, the proposed AGGAN is able to generate a dense 3D voxel grid of the face from a single unconstrained depth view. Specifically, AGGAN encodes the 3D facial geometry within a voxel space and utilizes an attention-guided GAN to model the illposed 2.5D depth-3D mapping. Multiple loss functions, which enforce the 3D facial geometry consistency, together with a prior distribution of facial surface points in voxel space are incorporated to guide the training process. Both qualitative and quantitative comparisons show that AGGAN recovers a more complete and smoother 3D facial shape, with the capability to handle a much wider range of view angles and resist to noise in the depth view than conventional methods

Jianwen Lou, Xiaoxu Cai, Junyu Dong et al.

We propose to learn a cascade of globally-optimized modular boosted ferns (GoMBF) to solve multi-modal facial motion regression for real-time 3D facial tracking from a monocular RGB camera. GoMBF is a deep composition of multiple regression models with each is a boosted ferns initially trained to predict partial motion parameters of the same modality, and then concatenated together via a global optimization step to form a singular strong boosted ferns that can effectively handle the whole regression target. It can explicitly cope with the modality variety in output variables, while manifesting increased fitting power and a faster learning speed comparing against the conventional boosted ferns. By further cascading a sequence of GoMBFs (GoMBF-Cascade) to regress facial motion parameters, we achieve competitive tracking performance on a variety of in-the-wild videos comparing to the state-of-the-art methods, which require much more training data or have higher computational complexity. It provides a robust and highly elegant solution to real-time 3D facial tracking using a small set of training data and hence makes it more practical in real-world applications.

Zhiming Chen, Kean Chen, Weiyao Lin et al.

Object detection using an oriented bounding box (OBB) can better target rotated objects by reducing the overlap with background areas. Existing OBB approaches are mostly built on horizontal bounding box detectors by introducing an additional angle dimension optimized by a distance loss. However, as the distance loss only minimizes the angle error of the OBB and that it loosely correlates to the IoU, it is insensitive to objects with high aspect ratios. Therefore, a novel loss, Pixels-IoU (PIoU) Loss, is formulated to exploit both the angle and IoU for accurate OBB regression. The PIoU loss is derived from IoU metric with a pixel-wise form, which is simple and suitable for both horizontal and oriented bounding box. To demonstrate its effectiveness, we evaluate the PIoU loss on both anchor-based and anchor-free frameworks. The experimental results show that PIoU loss can dramatically improve the performance of OBB detectors, particularly on objects with high aspect ratios and complex backgrounds. Besides, previous evaluation datasets did not include scenarios where the objects have high aspect ratios, hence a new dataset, Retail50K, is introduced to encourage the community to adapt OBB detectors for more complex environments.

Hui Yu

This paper presents a k-means-based multi-subpopulation particle swarm optimization, denoted as KMPSO, for training the neural network ensemble. In the proposed KMPSO, particles are dynamically partitioned into clusters via the k-means clustering algorithm at every iteration, and each of the resulting clusters is responsible for training a component neural network. The performance of the KMPSO has been evaluated on several benchmark problems. Our results show that the proposed method can effectively control the trade-off between the diversity and accuracy in the ensemble, thus achieving competitive results in comparison with related algorithms.

Jinguang Sun, Wanli Wang, Xian Wei et al.

The high dimensionality of hyperspectral images often results in the degradation of clustering performance. Due to the powerful ability of deep feature extraction and non-linear feature representation, the clustering algorithm based on deep learning has become a hot research topic in the field of hyperspectral remote sensing. However, most deep clustering algorithms for hyperspectral images utilize deep neural networks as feature extractor without considering prior knowledge constraints that are suitable for clustering. To solve this problem, we propose an intra-class distance constrained deep clustering algorithm for high-dimensional hyperspectral images. The proposed algorithm constrains the feature mapping procedure of the auto-encoder network by intra-class distance so that raw images are transformed from the original high-dimensional space to the low-dimensional feature space that is more conducive to clustering. Furthermore, the related learning process is treated as a joint optimization problem of deep feature extraction and clustering. Experimental results demonstrate the intense competitiveness of the proposed algorithm in comparison with state-of-the-art clustering methods of hyperspectral images.

Chao Gou, Tianyu Shen, Wenbo Zheng et al.

There has been much progress in data-driven artificial intelligence technology for medical image analysis in the last decades. However, it still remains challenging due to its distinctive complexity of acquiring and annotating image data, extracting medical domain knowledge, and explaining the diagnostic decision for medical image analysis. In this paper, we propose a data-knowledge-driven framework termed as Parallel Medical Imaging (PMI) for intelligent medical image analysis based on the methodology of interactive ACP-based parallel intelligence. In the PMI framework, computational experiments with predictive learning in a data-driven way are conducted to extract medical knowledge for diagnostic decision support. Artificial imaging systems are introduced to select and prescriptively generate medical image data in a knowledge-driven way to utilize medical domain knowledge. Through the closed-loop optimization based on parallel execution, our proposed PMI framework can boost the generalization ability and alleviate the limitation of medical interpretation for diagnostic decisions. Furthermore, we illustrate the preliminary implementation of PMI method through the case studies of mammogram analysis and skin lesion image analysis. Experimental results on several public medical image datasets demonstrate the effectiveness of proposed PMI.

Shu Zhang, Hui Yu, Ting Wang et al.

With the increasing demands of applications in virtual reality such as 3D films, virtual Human-Machine Interactions and virtual agents, the analysis of 3D human face analysis is considered to be more and more important as a fundamental step for those virtual reality tasks. Due to information provided by an additional dimension, 3D facial reconstruction enables aforementioned tasks to be achieved with higher accuracy than those based on 2D facial analysis. The denser the 3D facial model is, the more information it could provide. However, most existing dense 3D facial reconstruction methods require complicated processing and high system cost. To this end, this paper presents a novel method that simplifies the process of dense 3D facial reconstruction by employing only one frame of depth data obtained with an off-the-shelf RGB-D sensor. The experiments showed competitive results with real world data.

Hui Yu, Xiaofeng Wu, Wenbin Jiang et al.

The widely-used automatic evaluation metrics cannot adequately reflect the fluency of the translations. The n-gram-based metrics, like BLEU, limit the maximum length of matched fragments to n and cannot catch the matched fragments longer than n, so they can only reflect the fluency indirectly. METEOR, which is not limited by n-gram, uses the number of matched chunks but it does not consider the length of each chunk. In this paper, we propose an entropy-based method, which can sufficiently reflect the fluency of translations through the distribution of matched words. This method can easily combine with the widely-used automatic evaluation metrics to improve the evaluation of fluency. Experiments show that the correlations of BLEU and METEOR are improved on sentence level after combining with the entropy-based method on WMT 2010 and WMT 2012.

Hui Yu, Xiaofeng Wu, Wenbin Jiang et al.

Most of the syntax-based metrics obtain the similarity by comparing the sub-structures extracted from the trees of hypothesis and reference. These sub-structures are defined by human and can't express all the information in the trees because of the limited length of sub-structures. In addition, the overlapped parts between these sub-structures are computed repeatedly. To avoid these problems, we propose a novel automatic evaluation metric based on dependency parsing model, with no need to define sub-structures by human. First, we train a dependency parsing model by the reference dependency tree. Then we generate the hypothesis dependency tree and the corresponding probability by the dependency parsing model. The quality of the hypothesis can be judged by this probability. In order to obtain the lexicon similarity, we also introduce the unigram F-score to the new metric. Experiment results show that the new metric gets the state-of-the-art performance on system level, and is comparable with METEOR on sentence level.

Alexander Kadyrov, Hui Yu, Honghai Liu

There have been intensive research interests in ship detection and segmentation due to high demands on a wide range of civil applications in the last two decades. However, existing approaches, which are mainly based on statistical properties of images, fail to detect smaller ships and boats. Specifically, known techniques are not robust enough in view of inevitable small geometric and photometric changes in images consisting of ships. In this paper a novel approach for ship detection is proposed based on correlation of maritime images. The idea comes from the observation that a fine pattern of the sea surface changes considerably from time to time whereas the ship appearance basically keeps unchanged. We want to examine whether the images have a common unaltered part, a ship in this case. To this end, we developed a method - Focused Correlation (FC) to achieve robustness to geometric distortions of the image content. Various experiments have been conducted to evaluate the effectiveness of the proposed approach.