Jun Wan

Benjia Zhou, Zhigang Chen, Albert Clapés et al.

Sign Language Translation (SLT) is a challenging task due to its cross-domain nature, involving the translation of visual-gestural language to text. Many previous methods employ an intermediate representation, i.e., gloss sequences, to facilitate SLT, thus transforming it into a two-stage task of sign language recognition (SLR) followed by sign language translation (SLT). However, the scarcity of gloss-annotated sign language data, combined with the information bottleneck in the mid-level gloss representation, has hindered the further development of the SLT task. To address this challenge, we propose a novel Gloss-Free SLT based on Visual-Language Pretraining (GFSLT-VLP), which improves SLT by inheriting language-oriented prior knowledge from pre-trained models, without any gloss annotation assistance. Our approach involves two stages: (i) integrating Contrastive Language-Image Pre-training (CLIP) with masked self-supervised learning to create pre-tasks that bridge the semantic gap between visual and textual representations and restore masked sentences, and (ii) constructing an end-to-end architecture with an encoder-decoder-like structure that inherits the parameters of the pre-trained Visual Encoder and Text Decoder from the first stage. The seamless combination of these novel designs forms a robust sign language representation and significantly improves gloss-free sign language translation. In particular, we have achieved unprecedented improvements in terms of BLEU-4 score on the PHOENIX14T dataset (>+5) and the CSL-Daily dataset (>+3) compared to state-of-the-art gloss-free SLT methods. Furthermore, our approach also achieves competitive results on the PHOENIX14T dataset when compared with most of the gloss-based methods. Our code is available at https://github.com/zhoubenjia/GFSLT-VLP.

Benjia Zhou, Pichao Wang, Jun Wan et al.

Motion recognition is a promising direction in computer vision, but the training of video classification models is much harder than images due to insufficient data and considerable parameters. To get around this, some works strive to explore multimodal cues from RGB-D data. Although improving motion recognition to some extent, these methods still face sub-optimal situations in the following aspects: (i) Data augmentation, i.e., the scale of the RGB-D datasets is still limited, and few efforts have been made to explore novel data augmentation strategies for videos; (ii) Optimization mechanism, i.e., the tightly space-time-entangled network structure brings more challenges to spatiotemporal information modeling; And (iii) cross-modal knowledge fusion, i.e., the high similarity between multimodal representations caused to insufficient late fusion. To alleviate these drawbacks, we propose to improve RGB-D-based motion recognition both from data and algorithm perspectives in this paper. In more detail, firstly, we introduce a novel video data augmentation method dubbed ShuffleMix, which acts as a supplement to MixUp, to provide additional temporal regularization for motion recognition. Secondly, a Unified Multimodal De-coupling and multi-stage Re-coupling framework, termed UMDR, is proposed for video representation learning. Finally, a novel cross-modal Complement Feature Catcher (CFCer) is explored to mine potential commonalities features in multimodal information as the auxiliary fusion stream, to improve the late fusion results. The seamless combination of these novel designs forms a robust spatiotemporal representation and achieves better performance than state-of-the-art methods on four public motion datasets. Specifically, UMDR achieves unprecedented improvements of +4.5% on the Chalearn IsoGD dataset. Our code is available at https://github.com/zhoubenjia/MotionRGBD-PAMI.

Amazon AGI, Aaron Langford, Aayush Shah et al. · amazon-science

We present Amazon Nova, a new generation of state-of-the-art foundation models that deliver frontier intelligence and industry-leading price performance. Amazon Nova Pro is a highly-capable multimodal model with the best combination of accuracy, speed, and cost for a wide range of tasks. Amazon Nova Lite is a low-cost multimodal model that is lightning fast for processing images, video, documents and text. Amazon Nova Micro is a text-only model that delivers our lowest-latency responses at very low cost. Amazon Nova Canvas is an image generation model that creates professional grade images with rich customization controls. Amazon Nova Reel is a video generation model offering high-quality outputs, customization, and motion control. Our models were built responsibly and with a commitment to customer trust, security, and reliability. We report benchmarking results for core capabilities, agentic performance, long context, functional adaptation, runtime performance, and human evaluation.

Hao Fang, Ajian Liu, Jun Wan et al.

Face Anti-spoofing (FAS) is essential to secure face recognition systems from various physical attacks. However, recent research generally focuses on short-distance applications (i.e., phone unlocking) while lacking consideration of long-distance scenes (i.e., surveillance security checks). In order to promote relevant research and fill this gap in the community, we collect a large-scale Surveillance High-Fidelity Mask (SuHiFiMask) dataset captured under 40 surveillance scenes, which has 101 subjects from different age groups with 232 3D attacks (high-fidelity masks), 200 2D attacks (posters, portraits, and screens), and 2 adversarial attacks. In this scene, low image resolution and noise interference are new challenges faced in surveillance FAS. Together with the SuHiFiMask dataset, we propose a Contrastive Quality-Invariance Learning (CQIL) network to alleviate the performance degradation caused by image quality from three aspects: (1) An Image Quality Variable module (IQV) is introduced to recover image information associated with discrimination by combining the super-resolution network. (2) Using generated sample pairs to simulate quality variance distributions to help contrastive learning strategies obtain robust feature representation under quality variation. (3) A Separate Quality Network (SQN) is designed to learn discriminative features independent of image quality. Finally, a large number of experiments verify the quality of the SuHiFiMask dataset and the superiority of the proposed CQIL.

Zichang Tan, Jun Li, Jinhao Du et al.

Long-tailed visual recognition has received increasing attention in recent years. Due to the extremely imbalanced data distribution in long-tailed learning, the learning process shows great uncertainties. For example, the predictions of different experts on the same image vary remarkably despite the same training settings. To alleviate the uncertainty, we propose a Nested Collaborative Learning (NCL++) which tackles the long-tailed learning problem by a collaborative learning. To be specific, the collaborative learning consists of two folds, namely inter-expert collaborative learning (InterCL) and intra-expert collaborative learning (IntraCL). In-terCL learns multiple experts collaboratively and concurrently, aiming to transfer the knowledge among different experts. IntraCL is similar to InterCL, but it aims to conduct the collaborative learning on multiple augmented copies of the same image within the single expert. To achieve the collaborative learning in long-tailed learning, the balanced online distillation is proposed to force the consistent predictions among different experts and augmented copies, which reduces the learning uncertainties. Moreover, in order to improve the meticulous distinguishing ability on the confusing categories, we further propose a Hard Category Mining (HCM), which selects the negative categories with high predicted scores as the hard categories. Then, the collaborative learning is formulated in a nested way, in which the learning is conducted on not just all categories from a full perspective but some hard categories from a partial perspective. Extensive experiments manifest the superiority of our method with outperforming the state-of-the-art whether with using a single model or an ensemble. The code will be publicly released.

Jun Li, Zichang Tan, Jun Wan et al.

The networks trained on the long-tailed dataset vary remarkably, despite the same training settings, which shows the great uncertainty in long-tailed learning. To alleviate the uncertainty, we propose a Nested Collaborative Learning (NCL), which tackles the problem by collaboratively learning multiple experts together. NCL consists of two core components, namely Nested Individual Learning (NIL) and Nested Balanced Online Distillation (NBOD), which focus on the individual supervised learning for each single expert and the knowledge transferring among multiple experts, respectively. To learn representations more thoroughly, both NIL and NBOD are formulated in a nested way, in which the learning is conducted on not just all categories from a full perspective but some hard categories from a partial perspective. Regarding the learning in the partial perspective, we specifically select the negative categories with high predicted scores as the hard categories by using a proposed Hard Category Mining (HCM). In the NCL, the learning from two perspectives is nested, highly related and complementary, and helps the network to capture not only global and robust features but also meticulous distinguishing ability. Moreover, self-supervision is further utilized for feature enhancement. Extensive experiments manifest the superiority of our method with outperforming the state-of-the-art whether by using a single model or an ensemble.

Dong Wang, Jia Guo, Qiqi Shao et al.

Face anti-spoofing (FAS) is an essential mechanism for safeguarding the integrity of automated face recognition systems. Despite substantial advancements, the generalization of existing approaches to real-world applications remains challenging. This limitation can be attributed to the scarcity and lack of diversity in publicly available FAS datasets, which often leads to overfitting during training or saturation during testing. In terms of quantity, the number of spoof subjects is a critical determinant. Most datasets comprise fewer than 2,000 subjects. With regard to diversity, the majority of datasets consist of spoof samples collected in controlled environments using repetitive, mechanical processes. This data collection methodology results in homogenized samples and a dearth of scenario diversity. To address these shortcomings, we introduce the Wild Face Anti-Spoofing (WFAS) dataset, a large-scale, diverse FAS dataset collected in unconstrained settings. Our dataset encompasses 853,729 images of 321,751 spoof subjects and 529,571 images of 148,169 live subjects, representing a substantial increase in quantity. Moreover, our dataset incorporates spoof data obtained from the internet, spanning a wide array of scenarios and various commercial sensors, including 17 presentation attacks (PAs) that encompass both 2D and 3D forms. This novel data collection strategy markedly enhances FAS data diversity. Leveraging the WFAS dataset and Protocol 1 (Known-Type), we host the Wild Face Anti-Spoofing Challenge at the CVPR2023 workshop. Additionally, we meticulously evaluate representative methods using Protocol 1 and Protocol 2 (Unknown-Type). Through an in-depth examination of the challenge outcomes and benchmark baselines, we provide insightful analyses and propose potential avenues for future research. The dataset is released under Insightface.

Zenghao Bao, Zichang Tan, Jun Li et al.

Facial age estimation has received a lot of attention for its diverse application scenarios. Most existing studies treat each sample equally and aim to reduce the average estimation error for the entire dataset, which can be summarized as General Age Estimation. However, due to the long-tailed distribution prevalent in the dataset, treating all samples equally will inevitably bias the model toward the head classes (usually the adult with a majority of samples). Driven by this, some works suggest that each class should be treated equally to improve performance in tail classes (with a minority of samples), which can be summarized as Long-tailed Age Estimation. However, Long-tailed Age Estimation usually faces a performance trade-off, i.e., achieving improvement in tail classes by sacrificing the head classes. In this paper, our goal is to design a unified framework to perform well on both tasks, killing two birds with one stone. To this end, we propose a simple, effective, and flexible training paradigm named GLAE, which is two-fold. Our GLAE provides a surprising improvement on Morph II, reaching the lowest MAE and CMAE of 1.14 and 1.27 years, respectively. Compared to the previous best method, MAE dropped by up to 34%, which is an unprecedented improvement, and for the first time, MAE is close to 1 year old. Extensive experiments on other age benchmark datasets, including CACD, MIVIA, and Chalearn LAP 2015, also indicate that GLAE outperforms the state-of-the-art approaches significantly.

Hao Fang, Ajian Liu, Jun Wan et al.

Face Anti-spoofing (FAS) is essential to secure face recognition systems from various physical attacks. However, most of the studies lacked consideration of long-distance scenarios. Specifically, compared with FAS in traditional scenes such as phone unlocking, face payment, and self-service security inspection, FAS in long-distance such as station squares, parks, and self-service supermarkets are equally important, but it has not been sufficiently explored yet. In order to fill this gap in the FAS community, we collect a large-scale Surveillance High-Fidelity Mask (SuHiFiMask). SuHiFiMask contains $10,195$ videos from $101$ subjects of different age groups, which are collected by $7$ mainstream surveillance cameras. Based on this dataset and protocol-$3$ for evaluating the robustness of the algorithm under quality changes, we organized a face presentation attack detection challenge in surveillance scenarios. It attracted 180 teams for the development phase with a total of 37 teams qualifying for the final round. The organization team re-verified and re-ran the submitted code and used the results as the final ranking. In this paper, we present an overview of the challenge, including an introduction to the dataset used, the definition of the protocol, the evaluation metrics, and the announcement of the competition results. Finally, we present the top-ranked algorithms and the research ideas provided by the competition for attack detection in long-range surveillance scenarios.

Wenwen Min, Zhiceng Shi, Jun Zhang et al.

In recent years, the advent of spatial transcriptomics (ST) technology has unlocked unprecedented opportunities for delving into the complexities of gene expression patterns within intricate biological systems. Despite its transformative potential, the prohibitive cost of ST technology remains a significant barrier to its widespread adoption in large-scale studies. An alternative, more cost-effective strategy involves employing artificial intelligence to predict gene expression levels using readily accessible whole-slide images (WSIs) stained with Hematoxylin and Eosin (H\&E). However, existing methods have yet to fully capitalize on multimodal information provided by H&E images and ST data with spatial location. In this paper, we propose \textbf{mclSTExp}, a multimodal contrastive learning with Transformer and Densenet-121 encoder for Spatial Transcriptomics Expression prediction. We conceptualize each spot as a "word", integrating its intrinsic features with spatial context through the self-attention mechanism of a Transformer encoder. This integration is further enriched by incorporating image features via contrastive learning, thereby enhancing the predictive capability of our model. Our extensive evaluation of \textbf{mclSTExp} on two breast cancer datasets and a skin squamous cell carcinoma dataset demonstrates its superior performance in predicting spatial gene expression. Moreover, mclSTExp has shown promise in interpreting cancer-specific overexpressed genes, elucidating immune-related genes, and identifying specialized spatial domains annotated by pathologists. Our source code is available at https://github.com/shizhiceng/mclSTExp.

Jun Wan, Jun Liu, Jie Zhou et al.

Most facial landmark detection methods predict landmarks by mapping the input facial appearance features to landmark heatmaps and have achieved promising results. However, when the face image is suffering from large poses, heavy occlusions and complicated illuminations, they cannot learn discriminative feature representations and effective facial shape constraints, nor can they accurately predict the value of each element in the landmark heatmap, limiting their detection accuracy. To address this problem, we propose a novel Reference Heatmap Transformer (RHT) by introducing reference heatmap information for more precise facial landmark detection. The proposed RHT consists of a Soft Transformation Module (STM) and a Hard Transformation Module (HTM), which can cooperate with each other to encourage the accurate transformation of the reference heatmap information and facial shape constraints. Then, a Multi-Scale Feature Fusion Module (MSFFM) is proposed to fuse the transformed heatmap features and the semantic features learned from the original face images to enhance feature representations for producing more accurate target heatmaps. To the best of our knowledge, this is the first study to explore how to enhance facial landmark detection by transforming the reference heatmap information. The experimental results from challenging benchmark datasets demonstrate that our proposed method outperforms the state-of-the-art methods in the literature.

Hanshen Xiao, Jun Wan, Srinivas Devadas

Training large neural networks with meaningful/usable differential privacy security guarantees is a demanding challenge. In this paper, we tackle this problem by revisiting the two key operations in Differentially Private Stochastic Gradient Descent (DP-SGD): 1) iterative perturbation and 2) gradient clipping. We propose a generic optimization framework, called {\em ModelMix}, which performs random aggregation of intermediate model states. It strengthens the composite privacy analysis utilizing the entropy of the training trajectory and improves the $(ε, δ)$ DP security parameters by an order of magnitude. We provide rigorous analyses for both the utility guarantees and privacy amplification of ModelMix. In particular, we present a formal study on the effect of gradient clipping in DP-SGD, which provides theoretical instruction on how hyper-parameters should be selected. We also introduce a refined gradient clipping method, which can further sharpen the privacy loss in private learning when combined with ModelMix. Thorough experiments with significant privacy/utility improvement are presented to support our theory. We train a Resnet-20 network on CIFAR10 with $70.4\%$ accuracy via ModelMix given $(ε=8, δ=10^{-5})$ DP-budget, compared to the same performance but with $(ε=145.8,δ=10^{-5})$ using regular DP-SGD; assisted with additional public low-dimensional gradient embedding, one can further improve the accuracy to $79.1\%$ with $(ε=6.1, δ=10^{-5})$ DP-budget, compared to the same performance but with $(ε=111.2, δ=10^{-5})$ without ModelMix.

Benjia Zhou, Pichao Wang, Jun Wan et al.

Vision Transformers (ViTs) have shown promising performance compared with Convolutional Neural Networks (CNNs), but the training of ViTs is much harder than CNNs. In this paper, we define several metrics, including Dynamic Data Proportion (DDP) and Knowledge Assimilation Rate (KAR), to investigate the training process, and divide it into three periods accordingly: formation, growth and exploration. In particular, at the last stage of training, we observe that only a tiny portion of training examples is used to optimize the model. Given the data-hungry nature of ViTs, we thus ask a simple but important question: is it possible to provide abundant ``effective'' training examples at EVERY stage of training? To address this issue, we need to address two critical questions, \ie, how to measure the ``effectiveness'' of individual training examples, and how to systematically generate enough number of ``effective'' examples when they are running out. To answer the first question, we find that the ``difficulty'' of training samples can be adopted as an indicator to measure the ``effectiveness'' of training samples. To cope with the second question, we propose to dynamically adjust the ``difficulty'' distribution of the training data in these evolution stages. To achieve these two purposes, we propose a novel data-centric ViT training framework to dynamically measure the ``difficulty'' of training samples and generate ``effective'' samples for models at different training stages. Furthermore, to further enlarge the number of ``effective'' samples and alleviate the overfitting problem in the late training stage of ViTs, we propose a patch-level erasing strategy dubbed PatchErasing. Extensive experiments demonstrate the effectiveness of the proposed data-centric ViT training framework and techniques.

Senyuan Shi, Hao Tan, Zichang Tan et al.

The rapid evolution of generative models has precipitated a proliferation of fabricated content, posing significant challenges to existing Synthetic Image Detection (SID) methods. Capitalizing on advancements in vision-language models (e.g., CLIP), recent attempts have leveraged learnable textual prompts to identify synthetic images. However, they still leverage static prompt as a fixed boundary for real and fake images, failing to adapt to the varying types of forgery that emerge during inference. To overcome this issue, we propose **HydraPrompt**, an asymmetric prompting framework that dynamically adjusts the category centers by aligning with fine-grained image cues. Specifically, we propose an Asymmetric Prompt Adapter (**APA**): (1) for authentic category, we introduce a single set of prompts to capture the consistent representative patterns, which serves as a unified anchor for real content. While (2) for fake category, we construct sample-adaptive prompts that specialize in capturing diverse cues from different samples, enabling adaptive modeling of forgery image variations. To increase pronounced discriminability within different synthetic images, we further introduce a Conditional Supervised Contrastive (**CSC**) objective, which compacts the authentic representations while capturing fine-grained forgery clues. Extensive experiments on popular SID benchmarks demonstrate the state-of-the-art performance of our framework.

Hang Zou, Chenxi Du, Hui Zhang et al.

Facial recognition systems are susceptible to both physical and digital attacks, posing significant security risks. Traditional approaches often treat these two attack types separately due to their distinct characteristics. Thus, when being combined attacked, almost all methods could not deal. Some studies attempt to combine the sparse data from both types of attacks into a single dataset and try to find a common feature space, which is often impractical due to the space is difficult to be found or even non-existent. To overcome these challenges, we propose a novel approach that uses the sparse model to handle sparse data, utilizing different parameter groups to process distinct regions of the sparse feature space. Specifically, we employ the Mixture of Experts (MoE) framework in our model, expert parameters are matched to tokens with varying weights during training and adaptively activated during testing. However, the traditional MoE struggles with the complex and irregular classification boundaries of this problem. Thus, we introduce a flexible self-adapting weighting mechanism, enabling the model to better fit and adapt. In this paper, we proposed La-SoftMoE CLIP, which allows for more flexible adaptation to the Unified Attack Detection (UAD) task, significantly enhancing the model's capability to handle diversity attacks. Experiment results demonstrate that our proposed method has SOTA performance.

Hao Tan, Zichang Tan, Jun Li et al.

Multi-label image recognition is a fundamental task in computer vision. Recently, Vision-Language Models (VLMs) have made notable advancements in this area. However, previous methods fail to effectively leverage the rich knowledge in language models and often incorporate label semantics into visual features unidirectionally. To overcome these problems, we propose a Split-and-Synthesize Prompting with Gated Alignments (SSPA) framework to amplify the potential of VLMs. Specifically, we develop an in-context learning approach to associate the inherent knowledge from LLMs. Then we propose a novel Split-and-Synthesize Prompting (SSP) strategy to first model the generic knowledge and downstream label semantics individually and then aggregate them carefully through the quaternion network. Moreover, we present Gated Dual-Modal Alignments (GDMA) to bidirectionally interact visual and linguistic modalities while eliminating redundant cross-modal information, enabling more efficient region-level alignments. Rather than making the final prediction by a sharp manner in previous works, we propose a soft aggregator to jointly consider results from all image regions. With the help of flexible prompting and gated alignments, SSPA is generalizable to specific domains. Extensive experiments on nine datasets from three domains (i.e., natural, pedestrian attributes and remote sensing) demonstrate the state-of-the-art performance of SSPA. Further analyses verify the effectiveness of SSP and the interpretability of GDMA. The code will be made public.

Zhigang Chen, Benjia Zhou, Yiqing Huang et al.

Sign Language Representation Learning (SLRL) is crucial for a range of sign language-related downstream tasks such as Sign Language Translation (SLT) and Sign Language Retrieval (SLRet). Recently, many gloss-based and gloss-free SLRL methods have been proposed, showing promising performance. Among them, the gloss-free approach shows promise for strong scalability without relying on gloss annotations. However, it currently faces suboptimal solutions due to challenges in encoding the intricate, context-sensitive characteristics of sign language videos, mainly struggling to discern essential sign features using a non-monotonic video-text alignment strategy. Therefore, we introduce an innovative pretraining paradigm for gloss-free SLRL, called C${^2}$RL, in this paper. Specifically, rather than merely incorporating a non-monotonic semantic alignment of video and text to learn language-oriented sign features, we emphasize two pivotal aspects of SLRL: Implicit Content Learning (ICL) and Explicit Context Learning (ECL). ICL delves into the content of communication, capturing the nuances, emphasis, timing, and rhythm of the signs. In contrast, ECL focuses on understanding the contextual meaning of signs and converting them into equivalent sentences. Despite its simplicity, extensive experiments confirm that the joint optimization of ICL and ECL results in robust sign language representation and significant performance gains in gloss-free SLT and SLRet tasks. Notably, C${^2}$RL improves the BLEU-4 score by +5.3 on P14T, +10.6 on CSL-daily, +6.2 on OpenASL, and +1.3 on How2Sign. It also boosts the R@1 score by +8.3 on P14T, +14.4 on CSL-daily, and +5.9 on How2Sign. Additionally, we set a new baseline for the OpenASL dataset in the SLRet task.

Hao Tan, Jun Lan, Senyuan Shi et al.

The growing capability of video generation poses escalating security risks, making reliable detection increasingly essential. In this paper, we introduce VideoVeritas, a framework that integrates fine-grained perception and fact-based reasoning. We observe that while current multi-modal large language models (MLLMs) exhibit strong reasoning capacity, their granular perception ability remains limited. To mitigate this, we introduce Joint Preference Alignment and Perception Pretext Reinforcement Learning (PPRL). Specifically, rather than directly optimizing for detection task, we adopt general spatiotemporal grounding and self-supervised object counting in the RL stage, enhancing detection performance with simple perception pretext tasks. To facilitate robust evaluation, we further introduce MintVid, a light yet high-quality dataset containing 3K videos from 9 state-of-the-art generators, along with a real-world collected subset that has factual errors in content. Experimental results demonstrate that existing methods tend to bias towards either superficial reasoning or mechanical analysis, while VideoVeritas achieves more balanced performance across diverse benchmarks.

Hang Zou, Chenxi Du, Ajian Liu et al.

Iris recognition is widely used in high-security scenarios due to its stability and distinctiveness. However, iris images captured by different devices exhibit certain and device-related consistent differences, which has a greater impact on the classification algorithm for anti-spoofing. The iris of various races would also affect the classification, causing the risk of identity theft. So it is necessary to improve the cross-domain capabilities of the iris anti-spoofing (IAS) methods to enable it more robust in facing different races and devices. However, there is no existing protocol that is comprehensively available. To address this gap, we propose an Iris Anti-Spoofing Cross-Domain-Testing (IAS-CDT) Protocol, which involves 10 datasets, belonging to 7 databases, published by 4 institutions, and collected with 6 different devices. It contains three sub-protocols hierarchically, aimed at evaluating average performance, cross-racial generalization, and cross-device generalization of IAS models. Moreover, to address the cross-device generalization challenge brought by the IAS-CDT Protocol, we employ multiple model parameter sets to learn from the multiple sub-datasets. Specifically, we utilize the Mixture of Experts (MoE) to fit complex data distributions using multiple sub-neural networks. To further enhance the generalization capabilities, we propose a novel method Masked-MoE (MMoE), which randomly masks a portion of tokens for some experts and requires their outputs to be similar to the unmasked experts, which can effectively mitigate the overfitting issue of MoE. For the evaluation, we selected ResNet50, VIT-B/16, CLIP, and FLIP as representative models and benchmarked them under the proposed IAS-CDT Protocol.

Hao Tan, Jun Li, Yizhuang Zhou et al.

Vision-Language Models (VLMs) such as CLIP have demonstrated remarkable generalization capabilities to downstream tasks. However, existing prompt tuning based frameworks need to parallelize learnable textual inputs for all categories, suffering from massive GPU memory consumption when there is a large number of categories in the target dataset. Moreover, previous works require to include category names within prompts, exhibiting subpar performance when dealing with ambiguous category names. To address these shortcomings, we propose Compound Text-Guided Prompt Tuning (TGP-T) that significantly reduces resource demand while achieving superior performance. We introduce text supervision to the optimization of prompts, which enables two benefits: 1) releasing the model reliance on the pre-defined category names during inference, thereby enabling more flexible prompt generation; 2) reducing the number of inputs to the text encoder, which decreases GPU memory consumption significantly. Specifically, we found that compound text supervisions, i.e., category-wise and content-wise, is highly effective, since they provide inter-class separability and capture intra-class variations, respectively. Moreover, we condition the prompt generation on visual features through a module called Bonder, which facilitates the alignment between prompts and visual features. Extensive experiments on few-shot recognition and domain generalization demonstrate that TGP-T achieves superior performance with consistently lower training costs. It reduces GPU memory usage by 93% and attains a 2.5% performance gain on 16-shot ImageNet. The code is available at https://github.com/EricTan7/TGP-T.

Jun Wan, He Liu, Yujia Wu et al.

At present, deep neural network methods have played a dominant role in face alignment field. However, they generally use predefined network structures to predict landmarks, which tends to learn general features and leads to mediocre performance, e.g., they perform well on neutral samples but struggle with faces exhibiting large poses or occlusions. Moreover, they cannot effectively deal with semantic gaps and ambiguities among features at different scales, which may hinder them from learning efficient features. To address the above issues, in this paper, we propose a Dynamic Semantic-Aggregation Transformer (DSAT) for more discriminative and representative feature (i.e., specialized feature) learning. Specifically, a Dynamic Semantic-Aware (DSA) model is first proposed to partition samples into subsets and activate the specific pathways for them by estimating the semantic correlations of feature channels, making it possible to learn specialized features from each subset. Then, a novel Dynamic Semantic Specialization (DSS) model is designed to mine the homogeneous information from features at different scales for eliminating the semantic gap and ambiguities and enhancing the representation ability. Finally, by integrating the DSA model and DSS model into our proposed DSAT in both dynamic architecture and dynamic parameter manners, more specialized features can be learned for achieving more precise face alignment. It is interesting to show that harder samples can be handled by activating more feature channels. Extensive experiments on popular face alignment datasets demonstrate that our proposed DSAT outperforms state-of-the-art models in the literature.Our code is available at https://github.com/GERMINO-LiuHe/DSAT.

Zhi Rao, Yucheng Zhou, Benjia Zhou et al.

Gloss-free sign language translation (SLT) is hindered by two key challenges: **inadequate sign representation** that fails to capture nuanced visual cues, and **sentence-level semantic misalignment** in current LLM-based methods, which limits translation quality. To address these issues, we propose a three-stage **r**einforcing **v**ision-**l**anguage **f**ramework (**RVLF**). We build a large vision-language model (LVLM) specifically designed for sign language, and then combine it with reinforcement learning (RL) to adaptively enhance translation performance. First, for a sufficient representation of sign language, RVLF introduces an effective semantic representation learning mechanism that fuses skeleton-based motion cues with semantically rich visual features extracted via DINOv2, followed by instruction tuning to obtain a strong SLT-SFT baseline. Then, to improve sentence-level semantic misalignment, we introduce a GRPO-based optimization strategy that fine-tunes the SLT-SFT model with a reward function combining translation fidelity (BLEU) and sentence completeness (ROUGE), yielding the optimized model termed SLT-GRPO. Our conceptually simple framework yields substantial gains under the gloss-free SLT setting without pre-training on any external large-scale sign language datasets, improving BLEU-4 scores by +5.1, +1.11, +1.4, and +1.61 on the CSL-Daily, PHOENIX-2014T, How2Sign, and OpenASL datasets, respectively. To the best of our knowledge, this is the first work to incorporate GRPO into SLT. Extensive experiments and ablation studies validate the effectiveness of GRPO-based optimization in enhancing both translation quality and semantic consistency.

Kunchi Li, Jun Wan, Shan Yu

Rehearsal approaches in class incremental learning (CIL) suffer from decision boundary overfitting to new classes, which is mainly caused by two factors: insufficiency of old classes data for knowledge distillation and imbalanced data learning between the learned and new classes because of the limited storage memory. In this work, we present a simple but effective approach to tackle these two factors. First, we employ a re-sampling strategy and Mixup K}nowledge D}istillation (Re-MKD) to improve the performances of KD, which would greatly alleviate the overfitting problem. Specifically, we combine mixup and re-sampling strategies to synthesize adequate data used in KD training that are more consistent with the latent distribution between the learned and new classes. Second, we propose a novel incremental influence balance (IIB) method for CIL to tackle the classification of imbalanced data by extending the influence balance method into the CIL setting, which re-weights samples by their influences to create a proper decision boundary. With these two improvements, we present the effective decision boundary learning algorithm (EDBL) which improves the performance of KD and deals with the imbalanced data learning simultaneously. Experiments show that the proposed EDBL achieves state-of-the-art performances on several CIL benchmarks.

Shengkai Hu, Jiaqi Ma, Jun Wan et al.

All-in-One Image Restoration (AiOIR) aims to recover high-quality images from diverse degradations within a unified framework. However, existing methods often fail to explicitly model degradation types and struggle to adapt their restoration behavior to complex or mixed degradations. To address these issues, we propose ClusIR, a Cluster-Guided Image Restoration framework that explicitly models degradation semantics through learnable clustering and propagates cluster-aware cues across spatial and frequency domains for adaptive restoration. Specifically, ClusIR comprises two key components: a Probabilistic Cluster-Guided Routing Mechanism (PCGRM) and a Degradation-Aware Frequency Modulation Module (DAFMM). The proposed PCGRM disentangles degradation recognition from expert activation, enabling discriminative degradation perception and stable expert routing. Meanwhile, DAFMM leverages the cluster-guided priors to perform adaptive frequency decomposition and targeted modulation, collaboratively refining structural and textural representations for higher restoration fidelity. The cluster-guided synergy seamlessly bridges semantic cues with frequency-domain modulation, empowering ClusIR to attain remarkable restoration results across a wide range of degradations. Extensive experiments on diverse benchmarks validate that ClusIR reaches competitive performance under several scenarios.

Hao Tan, Zichang Tan, Jun Li et al.

Identifying multiple novel classes in an image, known as open-vocabulary multi-label recognition, is a challenging task in computer vision. Recent studies explore the transfer of powerful vision-language models such as CLIP. However, these approaches face two critical challenges: (1) The local semantics of CLIP are disrupted due to its global pre-training objectives, resulting in unreliable regional predictions. (2) The matching property between image regions and candidate labels has been neglected, relying instead on naive feature aggregation such as average pooling, which leads to spurious predictions from irrelevant regions. In this paper, we present RAM (Recover And Match), a novel framework that effectively addresses the above issues. To tackle the first problem, we propose Ladder Local Adapter (LLA) to enforce refocusing on local regions, recovering local semantics in a memory-friendly way. For the second issue, we propose Knowledge-Constrained Optimal Transport (KCOT) to suppress meaningless matching to non-GT labels by formulating the task as an optimal transport problem. As a result, RAM achieves state-of-the-art performance on various datasets from three distinct domains, and shows great potential to boost the existing methods. Code: https://github.com/EricTan7/RAM.

Jun Wan, Xinyu Xiong, Ning Chen et al.

Recently, deep learning based facial landmark detection (FLD) methods have achieved considerable success. However, in challenging scenarios such as large pose variations, illumination changes, and facial expression variations, they still struggle to accurately capture the geometric structure of the face, resulting in performance degradation. Moreover, the limited size and diversity of existing FLD datasets hinder robust model training, leading to reduced detection accuracy. To address these challenges, we propose a Frequency-Guided Task-Balancing Transformer (FGTBT), which enhances facial structure perception through frequency-domain modeling and multi-dataset unified training. Specifically, we propose a novel Fine-Grained Multi-Task Balancing loss (FMB-loss), which moves beyond coarse task-level balancing by assigning weights to individual landmarks based on their occurrence across datasets. This enables more effective unified training and mitigates the issue of inconsistent gradient magnitudes. Additionally, a Frequency-Guided Structure-Aware (FGSA) model is designed to utilize frequency-guided structure injection and regularization to help learn facial structure constraints. Extensive experimental results on popular benchmark datasets demonstrate that the integration of the proposed FMB-loss and FGSA model into our FGTBT framework achieves performance comparable to state-of-the-art methods. The code is available at https://github.com/Xi0ngxinyu/FGTBT.

Shengkai Hu, Haozhe Qi, Jun Wan et al.

Recent advances in deep learning have significantly improved facial landmark detection. However, existing facial landmark detection datasets often define different numbers of landmarks, and most mainstream methods can only be trained on a single dataset. This limits the model generalization to different datasets and hinders the development of a unified model. To address this issue, we propose Proto-Former, a unified, adaptive, end-to-end facial landmark detection framework that explicitly enhances dataset-specific facial structural representations (i.e., prototype). Proto-Former overcomes the limitations of single-dataset training by enabling joint training across multiple datasets within a unified architecture. Specifically, Proto-Former comprises two key components: an Adaptive Prototype-Aware Encoder (APAE) that performs adaptive feature extraction and learns prototype representations, and a Progressive Prototype-Aware Decoder (PPAD) that refines these prototypes to generate prompts that guide the model's attention to key facial regions. Furthermore, we introduce a novel Prototype-Aware (PA) loss, which achieves optimal path finding by constraining the selection weights of prototype experts. This loss function effectively resolves the problem of prototype expert addressing instability during multi-dataset training, alleviates gradient conflicts, and enables the extraction of more accurate facial structure features. Extensive experiments on widely used benchmark datasets demonstrate that our Proto-Former achieves superior performance compared to existing state-of-the-art methods. The code is publicly available at: https://github.com/Husk021118/Proto-Former.

Benjia Zhou, Pichao Wang, Jun Wan et al.

Decoupling spatiotemporal representation refers to decomposing the spatial and temporal features into dimension-independent factors. Although previous RGB-D-based motion recognition methods have achieved promising performance through the tightly coupled multi-modal spatiotemporal representation, they still suffer from (i) optimization difficulty under small data setting due to the tightly spatiotemporal-entangled modeling;(ii) information redundancy as it usually contains lots of marginal information that is weakly relevant to classification; and (iii) low interaction between multi-modal spatiotemporal information caused by insufficient late fusion. To alleviate these drawbacks, we propose to decouple and recouple spatiotemporal representation for RGB-D-based motion recognition. Specifically, we disentangle the task of learning spatiotemporal representation into 3 sub-tasks: (1) Learning high-quality and dimension independent features through a decoupled spatial and temporal modeling network. (2) Recoupling the decoupled representation to establish stronger space-time dependency. (3) Introducing a Cross-modal Adaptive Posterior Fusion (CAPF) mechanism to capture cross-modal spatiotemporal information from RGB-D data. Seamless combination of these novel designs forms a robust spatialtemporal representation and achieves better performance than state-of-the-art methods on four public motion datasets. Our code is available at https://github.com/damo-cv/MotionRGBD.

Benjia Zhou, Yunan Li, Jun Wan

Human gesture recognition has drawn much attention in the area of computer vision. However, the performance of gesture recognition is always influenced by some gesture-irrelevant factors like the background and the clothes of performers. Therefore, focusing on the regions of hand/arm is important to the gesture recognition. Meanwhile, a more adaptive architecture-searched network structure can also perform better than the block-fixed ones like Resnet since it increases the diversity of features in different stages of the network better. In this paper, we propose a regional attention with architecture-rebuilt 3D network (RAAR3DNet) for gesture recognition. We replace the fixed Inception modules with the automatically rebuilt structure through the network via Neural Architecture Search (NAS), owing to the different shape and representation ability of features in the early, middle, and late stage of the network. It enables the network to capture different levels of feature representations at different layers more adaptively. Meanwhile, we also design a stackable regional attention module called dynamic-static Attention (DSA), which derives a Gaussian guidance heatmap and dynamic motion map to highlight the hand/arm regions and the motion information in the spatial and temporal domains, respectively. Extensive experiments on two recent large-scale RGB-D gesture datasets validate the effectiveness of the proposed method and show it outperforms state-of-the-art methods. The codes of our method are available at: https://github.com/zhoubenjia/RAAR3DNet.

Jun Wan, Zhihui Lai, Jing Li et al.

Recently, heatmap regression has been widely explored in facial landmark detection and obtained remarkable performance. However, most of the existing heatmap regression-based facial landmark detection methods neglect to explore the high-order feature correlations, which is very important to learn more representative features and enhance shape constraints. Moreover, no explicit global shape constraints have been added to the final predicted landmarks, which leads to a reduction in accuracy. To address these issues, in this paper, we propose a Multi-order Multi-constraint Deep Network (MMDN) for more powerful feature correlations and shape constraints learning. Specifically, an Implicit Multi-order Correlating Geometry-aware (IMCG) model is proposed to introduce the multi-order spatial correlations and multi-order channel correlations for more discriminative representations. Furthermore, an Explicit Probability-based Boundary-adaptive Regression (EPBR) method is developed to enhance the global shape constraints and further search the semantically consistent landmarks in the predicted boundary for robust facial landmark detection. It's interesting to show that the proposed MMDN can generate more accurate boundary-adaptive landmark heatmaps and effectively enhance shape constraints to the predicted landmarks for faces with large pose variations and heavy occlusions. Experimental results on challenging benchmark datasets demonstrate the superiority of our MMDN over state-of-the-art facial landmark detection methods. The code has been publicly available at https://github.com/junwan2014/MMDN-master.

Zitong Yu, Benjia Zhou, Jun Wan et al.

Gesture recognition has attracted considerable attention owing to its great potential in applications. Although the great progress has been made recently in multi-modal learning methods, existing methods still lack effective integration to fully explore synergies among spatio-temporal modalities effectively for gesture recognition. The problems are partially due to the fact that the existing manually designed network architectures have low efficiency in the joint learning of multi-modalities. In this paper, we propose the first neural architecture search (NAS)-based method for RGB-D gesture recognition. The proposed method includes two key components: 1) enhanced temporal representation via the proposed 3D Central Difference Convolution (3D-CDC) family, which is able to capture rich temporal context via aggregating temporal difference information; and 2) optimized backbones for multi-sampling-rate branches and lateral connections among varied modalities. The resultant multi-modal multi-rate network provides a new perspective to understand the relationship between RGB and depth modalities and their temporal dynamics. Comprehensive experiments are performed on three benchmark datasets (IsoGD, NvGesture, and EgoGesture), demonstrating the state-of-the-art performance in both single- and multi-modality settings.The code is available at https://github.com/ZitongYu/3DCDC-NAS

Jiaqi Ma, Shengkai Hu, Xu Zhang et al.

All-in-One Image Restoration (AiOIR) tasks often involve diverse degradation that require robust and versatile strategies. However, most existing approaches typically lack explicit frequency modeling and rely on fixed or heuristic optimization schedules, which limit the generalization across heterogeneous degradation. To address these limitations, we propose EvoIR, an AiOIR-specific framework that introduces evolutionary frequency modulation for dynamic and adaptive image restoration. Specifically, EvoIR employs the Frequency-Modulated Module (FMM) that decomposes features into high- and low-frequency branches in an explicit manner and adaptively modulates them to enhance both structural fidelity and fine-grained details. Central to EvoIR, an Evolutionary Optimization Strategy (EOS) iteratively adjusts frequency-aware objectives through a population-based evolutionary process, dynamically balancing structural accuracy and perceptual fidelity. Its evolutionary guidance further mitigates gradient conflicts across degradation and accelerates convergence. By synergizing FMM and EOS, EvoIR yields greater improvements than using either component alone, underscoring their complementary roles. Extensive experiments on multiple benchmarks demonstrate that EvoIR outperforms state-of-the-art AiOIR methods.

Shuailin Xue, Jun Wan, Lihua Zhang et al.

Accurate detection of cancer tissue regions (CTR) enables deeper analysis of the tumor microenvironment and offers crucial insights into treatment response. Traditional CTR detection methods, which typically rely on the rich cellular morphology in histology images, are susceptible to a high rate of false positives due to morphological similarities across different tissue regions. The groundbreaking advances in spatial transcriptomics (ST) provide detailed cellular phenotypes and spatial localization information, offering new opportunities for more accurate cancer region detection. However, current methods are unable to effectively integrate histology images with ST data, especially in the context of cross-sample and cross-platform/batch settings for accomplishing the CTR detection. To address this challenge, we propose SpaCRD, a transfer learning-based method that deeply integrates histology images and ST data to enable reliable CTR detection across diverse samples, platforms, and batches. Once trained on source data, SpaCRD can be readily generalized to accurately detect cancerous regions across samples from different platforms and batches. The core of SpaCRD is a category-regularized variational reconstruction-guided bidirectional cross-attention fusion network, which enables the model to adaptively capture latent co-expression patterns between histological features and gene expression from multiple perspectives. Extensive benchmark analysis on 23 matched histology-ST datasets spanning various disease types, platforms, and batches demonstrates that SpaCRD consistently outperforms existing eight state-of-the-art methods in CTR detection.

Zhiceng Shi, Changmiao Wang, Jun Wan et al.

While spatial transcriptomics (ST) has advanced our understanding of gene expression in tissue context, its high experimental cost limits its large-scale application. Predicting ST from pathology images is a promising, cost-effective alternative, but existing methods struggle to capture complex cross-slide spatial relationships. To address the challenge, we propose SpaHGC, a multi-modal heterogeneous graph-based model that captures both intra-slice and inter-slice spot-spot relationships from histology images. It integrates local spatial context within the target slide and cross-slide similarities computed from image embeddings extracted by a pathology foundation model. These embeddings enable inter-slice knowledge transfer, and SpaHGC further incorporates Masked Graph Contrastive Learning to enhance feature representation and transfer spatial gene expression knowledge from reference to target slides, enabling it to model complex spatial dependencies and significantly improve prediction accuracy. We conducted comprehensive benchmarking on seven matched histology-ST datasets from different platforms, tissues, and cancer subtypes. The results demonstrate that SpaHGC significantly outperforms the existing nine state-of-the-art methods across all evaluation metrics. Additionally, the predictions are significantly enriched in multiple cancer-related pathways, thereby highlighting its strong biological relevance and application potential.

Ajian Liu, Shuai Xue, Jianwen Gan et al.

Domain generalization (DG) based Face Anti-Spoofing (FAS) aims to improve the model's performance on unseen domains. Existing methods either rely on domain labels to align domain-invariant feature spaces, or disentangle generalizable features from the whole sample, which inevitably lead to the distortion of semantic feature structures and achieve limited generalization. In this work, we make use of large-scale VLMs like CLIP and leverage the textual feature to dynamically adjust the classifier's weights for exploring generalizable visual features. Specifically, we propose a novel Class Free Prompt Learning (CFPL) paradigm for DG FAS, which utilizes two lightweight transformers, namely Content Q-Former (CQF) and Style Q-Former (SQF), to learn the different semantic prompts conditioned on content and style features by using a set of learnable query vectors, respectively. Thus, the generalizable prompt can be learned by two improvements: (1) A Prompt-Text Matched (PTM) supervision is introduced to ensure CQF learns visual representation that is most informative of the content description. (2) A Diversified Style Prompt (DSP) technology is proposed to diversify the learning of style prompts by mixing feature statistics between instance-specific styles. Finally, the learned text features modulate visual features to generalization through the designed Prompt Modulation (PM). Extensive experiments show that the CFPL is effective and outperforms the state-of-the-art methods on several cross-domain datasets.

Yichen Shi, Yuhao Gao, Yingxin Lai et al.

Multimodal large language models (MLLMs) have demonstrated strong capabilities in vision-related tasks, capitalizing on their visual semantic comprehension and reasoning capabilities. However, their ability to detect subtle visual spoofing and forgery clues in face attack detection tasks remains underexplored. In this paper, we introduce a benchmark, SHIELD, to evaluate MLLMs for face spoofing and forgery detection. Specifically, we design true/false and multiple-choice questions to assess MLLM performance on multimodal face data across two tasks. For the face anti-spoofing task, we evaluate three modalities (i.e., RGB, infrared, and depth) under six attack types. For the face forgery detection task, we evaluate GAN-based and diffusion-based data, incorporating visual and acoustic modalities. We conduct zero-shot and few-shot evaluations in standard and chain of thought (COT) settings. Additionally, we propose a novel multi-attribute chain of thought (MA-COT) paradigm for describing and judging various task-specific and task-irrelevant attributes of face images. The findings of this study demonstrate that MLLMs exhibit strong potential for addressing the challenges associated with the security of facial recognition technology applications.

Yucheng Zhou, Zhi Rao, Jun Wan et al.

Large Vision-Language Models (LVLMs) excel in cross-model tasks but experience performance declines in long-context reasoning due to overreliance on textual information and reduced visual dependency. In this study, we empirically analyze LVLMs in long-context reasoning, revealing that increased context length leads to a higher dependence on language at the expense of visual dependency. To address this issue, we propose a novel training-free context pruning method that selectively removes less critical textual information. Our approach enhances visual dependency and reduces textual noise, thereby improving LVLM performance in long-context reasoning. We validate our method by constructing a long-context dataset, demonstrating its effectiveness across various LVLMs. Moreover, further analysis confirms the robustness of different token pruning strategies and preliminary explores scaling laws between pruning rates and context length.

Hao Fang, Ajian Liu, Haocheng Yuan et al.

Face Recognition (FR) systems can suffer from physical (i.e., print photo) and digital (i.e., DeepFake) attacks. However, previous related work rarely considers both situations at the same time. This implies the deployment of multiple models and thus more computational burden. The main reasons for this lack of an integrated model are caused by two factors: (1) The lack of a dataset including both physical and digital attacks with ID consistency which means the same ID covers the real face and all attack types; (2) Given the large intra-class variance between these two attacks, it is difficult to learn a compact feature space to detect both attacks simultaneously. To address these issues, we collect a Unified physical-digital Attack dataset, called UniAttackData. The dataset consists of $1,800$ participations of 2 and 12 physical and digital attacks, respectively, resulting in a total of 29,706 videos. Then, we propose a Unified Attack Detection framework based on Vision-Language Models (VLMs), namely UniAttackDetection, which includes three main modules: the Teacher-Student Prompts (TSP) module, focused on acquiring unified and specific knowledge respectively; the Unified Knowledge Mining (UKM) module, designed to capture a comprehensive feature space; and the Sample-Level Prompt Interaction (SLPI) module, aimed at grasping sample-level semantics. These three modules seamlessly form a robust unified attack detection framework. Extensive experiments on UniAttackData and three other datasets demonstrate the superiority of our approach for unified face attack detection.

Lingrui Mei, Shenghua Liu, Yiwei Wang et al.

Large Language Models (LLMs) have made remarkable breakthroughs in reasoning, yet continue to struggle with hallucinations, logical errors, and inability to self-correct during complex multi-step tasks. Current approaches like chain-of-thought prompting offer limited reasoning capabilities that fail when precise step validation is required. We propose Environment Augmented Generation (EAG), a framework that enhances LLM reasoning through: (1) real-time environmental feedback validating each reasoning step, (2) dynamic branch exploration for investigating alternative solution paths when faced with errors, and (3) experience-based learning from successful reasoning trajectories. Unlike existing methods, EAG enables deliberate backtracking and strategic replanning through tight integration of execution feedback with branching exploration. Our a1-32B model achieves state-of-the-art performance among similar-sized models across all benchmarks, matching larger models like o1 on competition mathematics while outperforming comparable models by up to 24.4 percentage points. Analysis reveals EAG's distinctive scaling pattern: initial token investment in environment interaction yields substantial long-term performance dividends, with advantages amplifying proportionally to task complexity. EAG's theoretical framework demonstrates how environment interactivity and systematic branch exploration together establish a new paradigm for reliable machine reasoning, particularly for problems requiring precise multi-step calculation and logical verification.

Tianshun Han, Shengnan Gui, Yiqing Huang et al.

Speech-driven 3D facial animation has improved a lot recently while most related works only utilize acoustic modality and neglect the influence of visual and textual cues, leading to unsatisfactory results in terms of precision and coherence. We argue that visual and textual cues are not trivial information. Therefore, we present a novel framework, namely PMMTalk, using complementary Pseudo Multi-Modal features for improving the accuracy of facial animation. The framework entails three modules: PMMTalk encoder, cross-modal alignment module, and PMMTalk decoder. Specifically, the PMMTalk encoder employs the off-the-shelf talking head generation architecture and speech recognition technology to extract visual and textual information from speech, respectively. Subsequently, the cross-modal alignment module aligns the audio-image-text features at temporal and semantic levels. Then PMMTalk decoder is employed to predict lip-syncing facial blendshape coefficients. Contrary to prior methods, PMMTalk only requires an additional random reference face image but yields more accurate results. Additionally, it is artist-friendly as it seamlessly integrates into standard animation production workflows by introducing facial blendshape coefficients. Finally, given the scarcity of 3D talking face datasets, we introduce a large-scale 3D Chinese Audio-Visual Facial Animation (3D-CAVFA) dataset. Extensive experiments and user studies show that our approach outperforms the state of the art. We recommend watching the supplementary video.

Haocheng Yuan, Ajian Liu, Junze Zheng et al.

Face Anti-Spoofing (FAS) is crucial to safeguard Face Recognition (FR) Systems. In real-world scenarios, FRs are confronted with both physical and digital attacks. However, existing algorithms often address only one type of attack at a time, which poses significant limitations in real-world scenarios where FR systems face hybrid physical-digital threats. To facilitate the research of Unified Attack Detection (UAD) algorithms, a large-scale UniAttackData dataset has been collected. UniAttackData is the largest public dataset for Unified Attack Detection, with a total of 28,706 videos, where each unique identity encompasses all advanced attack types. Based on this dataset, we organized a Unified Physical-Digital Face Attack Detection Challenge to boost the research in Unified Attack Detections. It attracted 136 teams for the development phase, with 13 qualifying for the final round. The results re-verified by the organizing team were used for the final ranking. This paper comprehensively reviews the challenge, detailing the dataset introduction, protocol definition, evaluation criteria, and a summary of published results. Finally, we focus on the detailed analysis of the highest-performing algorithms and offer potential directions for unified physical-digital attack detection inspired by this competition. Challenge Website: https://sites.google.com/view/face-anti-spoofing-challenge/welcome/challengecvpr2024.

Yuhong He, Yongqi Zhang, Shizhu He et al.

Medical dialogue generation (MDG) has gained increasing attention due to its substantial practical value. Previous works typically employ a sequence-to-sequence framework to generate medical responses by modeling dialogue context as sequential text with annotated medical entities. While these methods have been successful in generating fluent responses, they fail to provide process explanations of reasoning and require extensive entity annotation. To address these limitations, we propose the method Bootstrap Prompting for Explicit Reasoning in MDG (BP4ER), which explicitly model MDG's multi-step reasoning process and iteratively enhance this reasoning process. We employ a least-to-most prompting strategy to guide a large language model (LLM) in explicit reasoning, breaking down MDG into simpler sub-questions. These sub-questions build on answers from previous ones. Additionally, we also introduce two distinct bootstrapping techniques for prompting, which autonomously correct errors and facilitate the LLM's explicit reasoning. This approach eliminates the need for entity annotation and increases the transparency of the MDG process by explicitly generating the intermediate reasoning chain. The experimental findings on the two public datasets indicate that BP4ER outperforms state-of-the-art methods in terms of both objective and subjective evaluation metrics.

Sunyuan Qiang, Yanyan Liang, Jun Wan et al.

Class-incremental learning (CIL) has emerged as a means to learn new classes incrementally without catastrophic forgetting of previous classes. Recently, CIL has undergone a paradigm shift towards dynamic architectures due to their superior performance. However, these models are still limited by the following aspects: (i) Data augmentation (DA), which are tightly coupled with CIL, remains under-explored in dynamic architecture scenarios. (ii) Feature representation. The discriminativeness of dynamic feature are sub-optimal and possess potential for refinement. (iii) Classifier. The misalignment between dynamic feature and classifier constrains the capabilities of the model. To tackle the aforementioned drawbacks, we propose the Dynamic Feature Learning and Matching (DFLM) model in this paper from above three perspectives. Specifically, we firstly introduce class weight information and non-stationary functions to extend the mix DA method for dynamically adjusting the focus on memory during training. Then, von Mises-Fisher (vMF) classifier is employed to effectively model the dynamic feature distribution and implicitly learn their discriminative properties. Finally, the matching loss is proposed to facilitate the alignment between the learned dynamic features and the classifier by minimizing the distribution distance. Extensive experiments on CIL benchmarks validate that our proposed model achieves significant performance improvements over existing methods.

Hao Tan, Zichang Tan, Jun Li et al.

Multi-label image recognition is a fundamental task in computer vision. Recently, vision-language models have made notable advancements in this area. However, previous methods often failed to effectively leverage the rich knowledge within language models and instead incorporated label semantics into visual features in a unidirectional manner. In this paper, we propose a Prompt-driven Visual-Linguistic Representation Learning (PVLR) framework to better leverage the capabilities of the linguistic modality. In PVLR, we first introduce a dual-prompting strategy comprising Knowledge-Aware Prompting (KAP) and Context-Aware Prompting (CAP). KAP utilizes fixed prompts to capture the intrinsic semantic knowledge and relationships across all labels, while CAP employs learnable prompts to capture context-aware label semantics and relationships. Later, we propose an Interaction and Fusion Module (IFM) to interact and fuse the representations obtained from KAP and CAP. In contrast to the unidirectional fusion in previous works, we introduce a Dual-Modal Attention (DMA) that enables bidirectional interaction between textual and visual features, yielding context-aware label representations and semantic-related visual representations, which are subsequently used to calculate similarities and generate final predictions for all labels. Extensive experiments on three popular datasets including MS-COCO, Pascal VOC 2007, and NUS-WIDE demonstrate the superiority of PVLR.

Shunxin Chen, Ajian Liu, Junze Zheng et al.

Facial recognition systems in real-world scenarios are susceptible to both digital and physical attacks. Previous methods have attempted to achieve classification by learning a comprehensive feature space. However, these methods have not adequately accounted for the inherent characteristics of physical and digital attack data, particularly the large intra class variation in attacks and the small inter-class variation between live and fake faces. To address these limitations, we propose the Fine-Grained MoE with Class-Aware Regularization CLIP framework (FG-MoE-CLIP-CAR), incorporating key improvements at both the feature and loss levels. At the feature level, we employ a Soft Mixture of Experts (Soft MoE) architecture to leverage different experts for specialized feature processing. Additionally, we refine the Soft MoE to capture more subtle differences among various types of fake faces. At the loss level, we introduce two constraint modules: the Disentanglement Module (DM) and the Cluster Distillation Module (CDM). The DM enhances class separability by increasing the distance between the centers of live and fake face classes. However, center-to-center constraints alone are insufficient to ensure distinctive representations for individual features. Thus, we propose the CDM to further cluster features around their respective class centers while maintaining separation from other classes. Moreover, specific attacks that significantly deviate from common attack patterns are often overlooked. To address this issue, our distance calculation prioritizes more distant features. Experimental results on two unified physical-digital attack datasets demonstrate that the proposed method achieves state-of-the-art (SOTA) performance.

Yongze Li, Ning Li, Ajian Liu et al.

Facial recognition systems are vulnerable to physical (e.g., printed photos) and digital (e.g., DeepFake) face attacks. Existing methods struggle to simultaneously detect physical and digital attacks due to: 1) significant intra-class variations between these attack types, and 2) the inadequacy of spatial information alone to comprehensively capture live and fake cues. To address these issues, we propose a unified attack detection model termed Frequency-Aware and Attack-Agnostic CLIP (FA\textsuperscript{3}-CLIP), which introduces attack-agnostic prompt learning to express generic live and fake cues derived from the fusion of spatial and frequency features, enabling unified detection of live faces and all categories of attacks. Specifically, the attack-agnostic prompt module generates generic live and fake prompts within the language branch to extract corresponding generic representations from both live and fake faces, guiding the model to learn a unified feature space for unified attack detection. Meanwhile, the module adaptively generates the live/fake conditional bias from the original spatial and frequency information to optimize the generic prompts accordingly, reducing the impact of intra-class variations. We further propose a dual-stream cues fusion framework in the vision branch, which leverages frequency information to complement subtle cues that are difficult to capture in the spatial domain. In addition, a frequency compression block is utilized in the frequency stream, which reduces redundancy in frequency features while preserving the diversity of crucial cues. We also establish new challenging protocols to facilitate unified face attack detection effectiveness. Experimental results demonstrate that the proposed method significantly improves performance in detecting physical and digital face attacks, achieving state-of-the-art results.

Hao Tan, Jun Lan, Zichang Tan et al.

Deepfake detection remains a formidable challenge due to the complex and evolving nature of fake content in real-world scenarios. However, existing academic benchmarks suffer from severe discrepancies from industrial practice, typically featuring homogeneous training sources and low-quality testing images, which hinder the practical deployments of current detectors. To mitigate this gap, we introduce HydraFake, a dataset that simulates real-world challenges with hierarchical generalization testing. Specifically, HydraFake involves diversified deepfake techniques and in-the-wild forgeries, along with rigorous training and evaluation protocol, covering unseen model architectures, emerging forgery techniques and novel data domains. Building on this resource, we propose Veritas, a multi-modal large language model (MLLM) based deepfake detector. Different from vanilla chain-of-thought (CoT), we introduce pattern-aware reasoning that involves critical reasoning patterns such as "planning" and "self-reflection" to emulate human forensic process. We further propose a two-stage training pipeline to seamlessly internalize such deepfake reasoning capacities into current MLLMs. Experiments on HydraFake dataset reveal that although previous detectors show great generalization on cross-model scenarios, they fall short on unseen forgeries and data domains. Our Veritas achieves significant gains across different OOD scenarios, and is capable of delivering transparent and faithful detection outputs.

Jun Wan

In 2020, OpenAI proposed the first type of Scaling Laws, describing the relationships between model loss and the scale of parameters, data, and training computation. In 2024, OpenAI proposed the second type of Scaling Laws, describing the relationship between model inference performance and inference computation. In this paper, we analyze LLMs training and inference processes from the perspective of lossless compression using conditional Kolmogorov complexity, and unify these two types of Scaling Laws. We find that both types of Scaling Laws improve approximation of conditional Kolmogorov complexity by increasing execution steps of Turing machine. The first type of Scaling Laws increases execution steps by increasing number of model parameters. The second type of Scaling Laws increases execution steps by increasing the number of intermediate tokens.

Lingrui Mei, Shenghua Liu, Yiwei Wang et al.

Long contexts challenge transformers: attention scores dilute across thousands of tokens, critical information is often lost in the middle, and models struggle to adapt to novel patterns at inference time. Recent work on test-time adaptation addresses this by maintaining a form of working memory -- transient parameters updated on the current context -- but existing approaches rely on uniform write policies that waste computation on low-utility regions and suffer from high gradient variance across semantically heterogeneous contexts. In this work, we reframe test-time adaptation as a budget-constrained memory consolidation problem, focusing on which parts of the context should be consolidated into working memory under limited computation. We propose Gdwm (Gated Differentiable Working Memory), a framework that introduces a write controller to gate the consolidation process. The controller estimates Contextual Utility, an information-theoretic measure of long-range contextual dependence, and allocates gradient steps accordingly while maintaining global coverage. Experiments on ZeroSCROLLS and LongBench v2 demonstrate that Gdwm achieves comparable or superior performance with 4$\times$ fewer gradient steps than uniform baselines, establishing a new efficiency-performance Pareto frontier for test-time adaptation.

Zheming Yang, Qi Guo, Jun Wan et al.

Multimodal large language models (MLLMs) enable powerful cross-modal reasoning capabilities but impose substantial computational and latency burdens, posing critical challenges for deployment on resource-constrained edge devices. In this paper, we propose MSAO, an adaptive modality sparsity-aware offloading framework with edge-cloud collaboration for efficient MLLM Inference. First, a lightweight heterogeneous modality-aware via fine-grained sparsity module performs spatial-temporal-modal joint analysis to compute the Modality Activation Sparsity (MAS) metric, which quantifies the necessity of each modality with minimal computational overhead. Second, an adaptive speculative edge-cloud collaborative offloading mechanism dynamically schedules workloads between edge and cloud based on the derived MAS scores and real-time system states, leveraging confidence-guided speculative execution to hide communication latency. Extensive experiments on VQAv2 and MMBench benchmarks demonstrate that MSAO achieves a 30% reduction in end-to-end latency and 30%-65% decrease in resource overhead, while delivering a throughput improvement of 1.5x to 2.3x compared to traditional approaches, all without compromising competitive accuracy.