Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Shihong LaoSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works243 from public data
- SiamRPN++: Evolution of Siamese Visual Tracking With Very Deep Networks2,477
This work proves the core reason Siamese trackers still have accuracy gap comes from the lack of strict translation invariance, and proposes a new model architecture to perform depth-wise and layer-wise aggregations, which not only improves the accuracy but also reduces the model size.
- Multiple object tracking: A literature review1,099
This work provides a thorough review on the development of this problem in recent decades and inspects the recent advances in various aspects and proposes some interesting directions for future research.
- An End-to-End Spatio-Temporal Attention Model for Human Action Recognition from Skeleton Data1,071
This work proposes an end-to-end spatial and temporal attention model for human action recognition from skeleton data on top of the Recurrent Neural Networks with Long Short-Term Memory (LSTM), which learns to selectively focus on discriminative joints of skeleton within each frame of the inputs and pays different levels of attention to the outputs of different frames.
- Co-Occurrence Feature Learning for Skeleton Based Action Recognition Using Regularized Deep LSTM Networks930
This work takes the skeleton as the input at each time slot and introduces a novel regularization scheme to learn the co-occurrence features of skeleton joints, and proposes a new dropout algorithm which simultaneously operates on the gates, cells, and output responses of the LSTM neurons.
- Pose-Driven Deep Convolutional Model for Person Re-identification843
A Pose-driven Deep Convolutional (PDC) model is proposed to learn improved feature extraction and matching models from end to end and explicitly leverages the human part cues to alleviate the pose variations and learn robust feature representations from both the global image and different local parts.
- The Visual Object Tracking VOT2016 Challenge Results776
The Visual Object Tracking challenge VOT2016 goes beyond its predecessors by introducing a new semi-automatic ground truth bounding box annotation methodology and extending the evaluation system with the no-reset experiment.
- View Adaptive Recurrent Neural Networks for High Performance Human Action Recognition from Skeleton Data675
A novel view adaptation scheme to automatically regulate observation viewpoints during the occurrence of an action by design a view adaptive recurrent neural network with LSTM architecture, which enables the network itself to adapt to the most suitable observation viewpoints from end to end.
- Semantics-Guided Neural Networks for Efficient Skeleton-Based Human Action Recognition645
A simple yet effective semantics-guided neural network (SGN) for skeleton-based action recognition that achieves the state-of-the-art performance on the NTU60, NTU120, and SYSU datasets.
- 585
- View Adaptive Neural Networks for High Performance Skeleton-Based Human Action Recognition522
A novel view adaptation scheme, which automatically determines the virtual observation viewpoints over the course of an action in a learning based data driven manner, and a two-stream scheme (referred to as VA-fusion) that fuses the scores of the two networks to provide the final prediction, obtaining enhanced performance.
- Transfer Learning Based Visual Tracking with Gaussian Processes Regression454
This paper directly analyze this probability of target appearance as exponentially related to the confidence of a classifier output using Gaussian Processes Regression (GPR), and introduces a latent variable to assist the tracking decision.
- 434
- 308
- 260
- 258
- 257
- 254
- 242
- Online Human Action Detection Using Joint Classification-Regression Recurrent Neural Networks226
This paper proposes a multi-task end-to-end Joint Classification-Regression Recurrent Neural Network to better explore the action type and temporal localization information and demonstrates very promising performance of this scheme.
- Robust Facial Landmark Detection via Recurrent Attentive-Refinement Networks217
A novel Recurrent Attentive-Refinement network for facial landmark detection under unconstrained conditions, suffering from challenges like facial occlusions and/or pose variations, which demonstrates superior performance in detecting challenging landmarks in comprehensive experiments and it also establishes new state-of-the-arts on the 300-W, COFW and AFLW benchmark datasets.
- Asymmetric 3D Convolutional Neural Networks for action recognition209
The asymmetric 3D-CNN model outperforms all the traditional 3D-CNN models in both effectiveness and efficiency, and its performance is comparable with that of recent state-of-the-art action recognition methods on both benchmarks.
- 205
- Selective Refinement Network for High Performance Face Detection171
A novel single-shot face detector, named Selective Refinement Network (SRN), which introduces novel two-step classification and regression operations selectively into an anchor-based face detector to reduce false positives and improve location accuracy simultaneously.
- 137
- 136
- Contrastive Context-Aware Learning for 3D High-Fidelity Mask Face Presentation Attack Detection130
This work introduces a large-scale High-Fidelity Mask dataset, namely HiFiMask, and proposes a novel Contrastive Context-aware Learning (CCL) framework for supervised PAD tasks, which is able to learn by leveraging rich contexts accurately among pairs of live faces and high-fidelity mask attacks.
- DriveWorld: 4D Pre-Trained Scene Understanding via World Models for Autonomous Driving129
This paper proposes a Memory State-Space Model for spatiotemporal modelling, which consists of a Dynamic Memory Bank module for learning temporal-aware latent dynamics to predict future changes and a Static Scene Propagation module for learning spatial-aware latent statics to offer comprehensive scene contexts.
- PedHunter: Occlusion Robust Pedestrian Detector in Crowded Scenes121
The proposed method, namely PedHunter, introduces strong occlusion handling ability to existing region-based detection networks without bringing extra computations in the inference stage, and designs a mask-guided module to leverage the head information to enhance the feature representation learning of the backbone network.
- Diagnosing deep learning models for high accuracy age estimation from a single image112
A deep multi-task age estimation model is built which obtains a MAE of 2.96 on the Morph II dataset and 5.75 on the WebFace dataset, both of which improve previous best results by a large margin.
- 105
- 103
- DiffCR: A Fast Conditional Diffusion Framework for Cloud Removal From Optical Satellite Images102
A novel framework called DiffCR is presented, which leverages conditional guided diffusion with deep convolutional networks for high-performance cloud removal for optical satellite imagery and introduces a decoupled encoder for conditional image feature extraction and proposes a novel and efficient time and condition fusion block within the cloud removal model.
- "Wow! you are so beautiful today!"97
A multiple tree-structured super-graphs model is proposed to explore the complex relationships among the high-level beauty attributes, mid-level beauty-related attributes and low-level image features and then based on this model, the most compatible beauty attributes for a given facial image can be efficiently inferred.
- 3D-Aided Deep Pose-Invariant Face Recognition83
A 3D-Aided Deep Pose-Invariant Face Recognition Model (3D-PIM), which automatically recovers realistic frontal faces from arbitrary poses through a 3D face model in a novel way, and demonstrates superiority of the proposed model over state-of-the-arts.
- 83
- Pose Partition Networks for Multi-person Pose Estimation75
This paper proposes a novel Pose Partition Network (PPN) to address the challenging multi-person pose estimation problem and implements PPN with the Hourglass architecture as the backbone network to simultaneously learn joint detector and dense regressor.
- Domain Adaptive Attention Learning for Unsupervised Person Re-Identification74
A domain adaptive attention learning approach to reliably transfer discriminative representation from the labeled source domain to the unlabeled target domain is proposed and outperforms the state-of-the-arts.
- AlphaHoldem: High-Performance Artificial Intelligence for Heads-Up No-Limit Poker via End-to-End Reinforcement Learning72
This work presents AlphaHoldem, a high-performance and lightweight HUNL AI obtained with an end-to-end self-play reinforcement learning framework that adopts a pseudo-siamese architecture to directly learn from the input state information to the output actions by competing the learned model with its different historical versions.
- Human Pose Estimation Using Global and Local Normalization71
A two-stage normalization scheme, human body normalization and limb normalization, is presented to make the distribution of the relative joint locations compact, resulting in easier learning of convolutional spatial models and more accurate pose estimation.
- Multi-type attributes driven multi-camera person re-identification69
A weakly supervised multi-type attribute learning framework which considers the contextual cues among attributes and progressively boosts the accuracy of attributes only using a limited number of labeled data is proposed.
- 69
- 68
- 65
- RDSNet: A New Deep Architecture forReciprocal Object Detection and Instance Segmentation65
RDSNet is presented, a novel deep architecture for reciprocal object detection and instance segmentation that combines a correlation module and a cropping module to yield instance masks, as well as a mask based boundary refinement module for more accurate bounding boxes.
- Relational Learning for Joint Head and Human Detection64
This work designs a head-body relationship discriminating module to perform relational learning between heads and human bodies, and leverage this learned relationship to regain the suppressed human detections and reduce head false positives.
- Temporal–Spatial Mapping for Action Recognition64
This work introduces a simple yet effective operation, termed temporal–spatial mapping, for capturing the temporal evolution of the frames by jointly analyzing all the frames of a video, and proposes a temporal attention model within a shallow convolutional neural network to efficiently exploit the temporal-spatial dynamics.
- 64
- Multi-Cue Illumination Estimation via a Tree-Structured Group Joint Sparse Representation64
Tests show that the proposed multi-cue illumination estimation method works better than existing methods, most of which are based on using only a single cue type, for example, a binarized color histogram or simple image statistic such as the mean RGB.
- 62
- Visual Tracking via Spatially Aligned Correlation Filters Network57
An end-to-end deep architecture is proposed to incorporate geometric transformations into a correlation filters based network, which introduces a novel spatial alignment module that provides continuous feedback for transforming the target from the border to the center with a normalized aspect ratio.
- Joint Scale-Spatial Correlation Tracking with Adaptive Rotation Estimation57
This work presents a fully functional correlation filter based tracking algorithm which is able to simultaneously model target appearance changes from spatial displacements, scale variations, and rotation transformations, and is efficiently and effectively performed in the joint space with fast Fourier Transform.
- D2C: Deep cumulatively and comparatively learning for human age estimation56
This work uses a convolutional neural network which can learn discriminative aging features from raw face images without any handcrafting and proposes a novel cumulative hidden layer which is supervised by a point-wise cumulative signal to combat the sample imbalance problem.
- Multi-Prototype Networks for Unconstrained Set-based Face Recognition52
A novel Multi-Prototype Network (MPNet) model is proposed to learn multiple prototype face representations adaptively from the media sets to demonstrate superiority of the proposed model over state-of-the-arts.
- 52
- 47
- 46
- Towards human-like and transhuman perception in AI 2.0: a review46
This paper briefly review the state-of-the-art advances across different areas of perception, including visual perception, auditory perception, speech perception, and perceptual information processing and learning engines and envision several R&D trends in intelligent perception for the forthcoming era of AI 2.0.
- Semi-Supervised Tensor-Based Graph Embedding Learning and Its Application to Visual Discriminant Tracking45
This paper treats an image patch as a 2-order tensor which preserves the original image structure and proposes a transfer-learningbased semi-supervised strategy to iteratively adjust the embedding space into which discriminative information obtained from earlier times is transferred.
- 45
- 42
- 41
- Do not Lose the Details: Reinforced Representation Learning for High Performance Visual Tracking38
This work presents a novel end-to-end trainable CNN model for high performance visual object tracking that learns both low-level fine-grained representations and a high-level semantic embedding space in a mutual reinforced way, and a multi-task learning strategy is proposed to perform the correlation analysis on representations from both levels.
- 38
- Recognizing Profile Faces by Imagining Frontal View37
Qualitative and quantitative experiments on both controlled and in-the-wild benchmark datasets demonstrate the superiority of the proposed Pose-Invariant Model (PIM) for face recognition in the wild, with three distinct novelties.
- 36
- 36
- 34
- Speech-Driven 3D Face Animation with Composite and Regional Facial Movements34
This paper introduces an adaptive modulation module that employs arbitrary facial movements to dynamically adjust speech-driven facial movements across frames on a global scale and presents a non-autoregressive backbone for translating audio to 3D facial movements, which maintains high-frequency nuances of facial movements and facilitates efficient inference.
- 34
- 33
- The 3rd Anti-UAV Workshop & Challenge: Methods and Results32
A brief summary of the 3rd Anti-UAV Workshop&Challenge is provided including brief introductions to the top three methods in each track including brief introduction to the bottom three methods.
- Evidential Detection and Tracking Collaboration: New Problem, Benchmark and Algorithm for Robust Anti-UAV System32
Extensive experiments show the method outperforms SOTA approaches and validate the ability of AntiUAV600 to enhance UAV perception performance due to its large scale and complexity.
- 32
- 32
- GroupDancer: Music to Multi-People Dance Synthesis with Style Collaboration30
Comprehensive evaluations on the MDC dataset demonstrate that the proposed GroupDancer model can synthesize quite satisfactory group dance synthesis results with style collaboration, and a novel Music-driven Group Dance Synthesis task is introduced.
- 30
- 30
- Asymmetric interaction preference induces cooperation in human-agent hybrid game29
This study proposes a human-agent hybrid prisoner’s dilemma game system within the framework of evolutionary game theory, and defines subjective preferences for individuals in various groups, including interaction preferences between homogeneous and heterogeneous groups.
- 28
- SDDM: Score-Decomposed Diffusion Models on Manifolds for Unpaired Image-to-Image Translation26
This work presents a new score-decomposed diffusion model (SDDM) on manifolds to explicitly optimize the tangled distributions during image generation to outperforms existing SBDM-based methods with much fewer diffusion steps on several I2I benchmarks.
- Towards stabilizing facial landmark detection and tracking via hierarchical filtering: A new method25
This work presents a novel hierarchical filtering method for stabilized facial landmark detection and tracking in video frames, which guarantees the robustness of global whole facial shape tracking and the adaptivity of local facial parts tracking.
- Facilitating cooperation in human-agent hybrid populations through autonomous agents24
Focusing on well-mixed populations, it is found that cooperative AAs have a limited impact in the prisoner’s dilemma games but facilitate cooperation in the stag hunt games, and imitation strength and population structure are critical for the emergence of human cooperation in human-agent hybrid populations.
- Distractor-aware discrimination learning for online multiple object tracking24
This work proposes a method for learning a distractor-aware discriminative model that can handle continuous missed and inaccurate detection problems due to the occlusion or the motion blur and proposes a multi-stage tracking pipeline designed for automatic trajectory initialization,propagation, and termination.
- Rank-1 Tensor Approximation for High-Order Association in Multi-target Tracking24
A tensor power iteration algorithm to effectively capture high-order motion information as well as appearance variation in multi-target tracking by reshaping MTT as a rank-1 tensor approximation problem (R1TA).
- BodyGen: Advancing Towards Efficient Embodiment Co-Design23
BodyGen is proposed, which utilizes topology-aware self-attention for both design and control, enabling efficient morphology representation with lightweight model sizes and a temporal credit assignment mechanism that ensures balanced reward signals for optimization.
- MMTSA22
An efficient multimodal neural architecture for HAR using an RGB camera and inertial measurement units (IMUs) and an inter-segment attention module called Multimodal Temporal Segment Attention Network (MMTSA) is proposed.
- 21
- DecenterNet: Bottom-Up Human Pose Estimation Via Decentralized Pose Representation19
The proposed DecenterNet is an end-to-end deep architecture to perform robust and efficient pose estimation in crowded scenes and introduces a decentralized pose representation that uses all visible keypoints as the root points to represent human poses, which is more robust in the entanglement area.
- 19
- 19
- 18
- Dynamic Dictionary Learning for Remote Sensing Image Segmentation17
A dynamic dictionary learning framework that explicitly models class ID embeddings through iterative refinement is introduced, ensuring compact intra-class distributions while maximizing inter-class separability and a contrastive constraint is applied to the dictionary space.
- 17
- 17
- 17
- 16
- 16
- AutoCFR: Learning to Design Counterfactual Regret Minimization Algorithms15
This work designs a search language that is rich enough to represent many existing hand-designed CFR variants and exploits a scalable regularized evolution algorithm with a bag of acceleration techniques to efficiently search over the combinatorial space of algorithms defined by this language.
- 14
- Adapting Vision Foundation Models for Robust Cloud Segmentation in Remote Sensing Images14
This article presents a parameter-efficient adaptive approach, termed Cloud-Adapter, designed to enhance the accuracy and robustness of cloud segmentation, which leverages a VFM pretrained on general domain data, which remains frozen, eliminating the need for additional training.
- GTR: Guided Thought Reinforcement Prevents Thought Collapse in RL-Based VLM Agent Training14
This work investigates thought collapse through extensive experiments on complex card games, such as 24 points, and embodied tasks from ALFWorld, and proposes an automated corrector that evaluates and refines the agent's reasoning at each $R L$ step to counteract thought collapse.
- UV-Mamba: A DCN-Enhanced State Space Model for Urban Village Boundary Identification in High-Resolution Remote Sensing Images14
A novel and efficient neural network model called UV-Mamba for accurate boundary detection in high-resolution remote sensing images that mitigates the memory loss problem in lengthy sequence modeling, which arises in state space models (SSM) with increasing image size, by incorporating deformable convolutions (DCN).
- UniParser: Multi-Human Parsing With Unified Correlation Representation Learning14
By unifying instance-level and category-level output, UniParser circumvents manually designed post-processing techniques and surpasses state-of-the-art methods, achieving 49.3% AP on MHPv2.0 and 60.4% AP on CIHP.
- Single-Stage Multi-human Parsing via Point Sets and Center-based Offsets14
A high-performance Single-stage Multi-human Parsing (SMP) deep architecture that decouples the multi-human parsing problem into two fine-grained sub-problems,i.e., locating the human body and parts.
- MMFace4D: A Large-Scale Multi-Modal 4D Face Dataset for Audio-Driven 3D Face Animation13
This work proposes MMFace4D, a large-scale multi-modal 4D (3D sequence) face dataset consisting of 431 identities, 35,904 sequences, and 3.9 million frames, and constructs a non-autoregressive framework for audio-driven 3D face animation that surpasses contemporary state-of-the-art approaches both qualitatively and quantitatively.
- 13
- 13
- Circle & Search13
An innovative product retrieval system, called Circle & Search, enables users to naturally indicate any preferred product by simply circling the product in images as the visual query, and then returns visually and semantically similar products to the users.
- 13
- Dual L1-Normalized Context Aware Tensor Power Iteration and Its Applications to Multi-object Tracking and Multi-graph Matching12
In this paper, multi-dimensional assignment is formulated as a rank-1 tensor approximation problem and a dual L1-normalized context/hyper-context aware tensor power iteration optimization method is proposed, applied to multi-object tracking and multi-graph matching.
- 12
- 12
- StickMotion : Generating 3D Human Motions by Drawing a Stickman11
StickMotion is introduced, an efficient diffusion-based network designed for multi-condition scenarios, which generates desired motions based on traditional text and the authors' proposed stickman conditions for global and local control of these motions, respectively.
- Towards Robust Pansharpening: A Large-Scale High-Resolution Multi-Scene Dataset and Novel Approach11
PanBench, a high-resolution multi-scene dataset containing all mainstream satellites and comprising 5898 pairs of samples is introduced, and a Cascaded Multiscale Fusion Network (CMFNet) for pansharpening is proposed to avoid irreversible loss of spectral information.
- High performance person re-identification via a boosting ranking ensemble11
This paper proposes a novel framework for person re-identification called Boosting Ranking Ensemble (BRE), which adaptively assembles features and metrics from a ranking perspective and explores an effective rectifier loss in the proposed BRE framework to alleviate the influence of the misclassified samples.
- 11
- SERVE: Soft and Equalized Residual VEctors for image retrieval11
This work proposes a novel image signature termed Soft and Equalized Residual VEctors (SERVE) which is more discriminatively formulated and maintains higher accuracy and outperforms existing state-of-the-art methods and provides a promising paradigm for the image retrieval task.
- 10
- 10
- 10
- 10
- Translating Motion to Notation: Hand Labanotation for Intuitive and Comprehensive Hand Movement Documentation9
A novel notation system, named Hand Labanotation (HL), for hand movement documentation, and a novel multi-view transformer architecture for automatically translating hand movements to HL, which makes the method a general tool for hand movement documentation.
- 9
- Spatial Preserved Graph Convolution Networks for Person Re-identification9
A graph convolution network based model for person re-identification is proposed to learn more discriminative feature embeddings, where a graph-structured relationship between person images and person parts are together integrated.
- 9
- Opponent Modeling with In-context Search8
A novel approach based on in-context learning and decision-time search named OMIS that converges in opponent policy recognition and has good generalization properties and demonstrates more effective and stable adaptation to opponents than other approaches.
- Advancing Video Synchronization with Fractional Frame Analysis: Introducing a Novel Dataset and Model8
This work proposes an Inter-Frame and Intra-Frame Desynchronized Dataset (IFID), which labels fractional time intervals between two video clips, and develops a novel model based on the Transformer architecture, named InSynFormer, for synchronizing inter-frame and intra-frame.
- The Exploration-Exploitation Dilemma Revisited: An Entropy Perspective8
An end-to-end adaptive framework called AdaZero is established, which automatically determines whether to explore or to exploit as well as their balance of strength, by revealing the relationship between entropy and the dynamic adaptive process of exploration and exploitation.
- Transductive Off-policy Proximal Policy Optimization8
This paper introduces a novel off-policy extension to the original PPO method, christened Transductive Off-policy PPO (ToPPO), and provides theoretical justification for incorporating off-policy data in PPO training and prudent guidelines for its safe application.
- 8
- 8
- 8
- Habitizing Diffusion Planning for Efficient and Effective Decision Making7
Habi is introduced, a general framework that transforms powerful but slow diffusion planning models into fast decision-making models, which mimics the cognitive process in the brain that costly goal-directed behavior gradually transitions to efficient habitual behavior with repetitive practice.
- An Open-Ended Learning Framework for Opponent Modeling7
A novel Open-Ended Opponent Modeling (OEOM) framework, which continuously generates opponents with diverse strengths and styles to reduce the possibility of OOD situations occurring during testing, and a simple yet effective OM approach that naturally fits within the OEOM framework.
- 7
- SkatingVerse: A large‐scale benchmark for comprehensive evaluation on human action understanding7
SkatingVerse significantly helps the training and evaluation of HAU methods, the performance of existing HAU methods has much room to improve, and SkatingVerse helps to reduce such gaps, and unifying relevant tasks in HAU through a uniform dataset can facilitate more practical applications.
- Complex pathways to cooperation emergent from asymmetry in heterogeneous populations7
The results indicate that in an infinite well-mixed population, defection remains the dominant strategy when strong players engage in the prisoner's dilemma game, however, if strong players play snowdrift games, global cooperation increases with the proportion of strong players.
- 7
- Versatile Face Animator: Driving Arbitrary 3D Facial Avatar in RGBD Space7
A novel self-supervised framework, Versatile Face Animator, which combines facial motion capture with motion retargeting in an end-to-end manner, eliminating the need for blendshapes or rigs is proposed.
- Exploration via State influence Modeling7
This work introduces a general intrinsic reward construction method to evaluate the social influence of states dynamically and shows its high exploration efficiency in hard exploration tasks with sparse rewards.
- Intelligent technologies of human-computer gaming7
This work analyzes the research status and dissects the key elements and intrinsic gaming mechanisms of human-computer gaming, and proposes a game learning-based theoretical research framework for HCG.
- OpenHoldem: An Open Toolkit for Large-Scale Imperfect-Information Game Research.7
OpenHoldem is presented, an integrated toolkit for large-scale imperfect-information game research using NLTH and makes three main contributions to this research direction: a standardized evaluation protocol for thoroughly evaluating different NLTH AIs, three publicly available strong baselines for NLTH AI, and an online testing platform with easy-to-use APIs for public NLTHAI evaluation.
- 7
- 7
- 7
- MMTSA: Multimodal Temporal Segment Attention Network for Efficient Human Activity Recognition6
This paper proposes a new novel multimodal neural architecture based on RGB and IMU wearable sensors for human activity recognition called Multimodal Temporal Segment Attention Network (MMTSA), which first employs a multimodAL data isomorphism mechanism based on Gramian Angular Field and then applies a novel multi-modal sparse sampling method to reduce redundancy.
- Real-Time Deep Video SpaTial Resolution UpConversion SysTem (STRUCT++ Demo)6
This work presents a real-time deep video SpaTial Resolution UpConversion SysTem (STRUCT++), built on a Global context aggregation and Local queue jumping Network (GLNet), which achieves state-of-the-art performance for real- time video SR.
- 6
- Multi-View 3D Human Pose Estimation with Weakly Synchronized Images5
SyncDiffPose is proposed, a novel model based on the diffusion method for pose estimation to denoise the error in weakly synchronized images from multiple views, which can perform pose estimation without hardware calibration.
- 5
- High-Fidelity Lake Extraction Via Two-Stage Prompt Enhancement: Establishing A Novel Baseline and Benchmark5
A prompt-based dataset construction approach that provides approximate lake locations using point, box, and mask prompts, and a two-stage prompt enhancement framework, LEPrompter, with prompt-based and prompt-free stages during training.
- Semantics2Hands: Transferring Hand Motion Semantics between Avatars5
This paper introduces a novel anatomy-based semantic matrix (ASM) that encodes the semantics of hand motions and obtains a mapping function from the source ASM to the target hand joint rotations by employing an anatomy- based semantics reconstruction network (ASRN).
- 5
- Hybrid Learning for Multi-agent Cooperation with Sub-optimal Demonstrations5
Experimental results show that the approach outperforms many competing demonstration-based methods and is evaluated on a real-time strategy combat game.
- Spatial alignment network for facial landmark localization5
A Spatial Alignment Network (SAN), which consist of two modules, like the transformation sub- network and the estimation sub-network, which achieves good performance on several public challenging datasets.
- Scene Aware Detection and Block Assignment Tracking in crowded scenes5
The SAD and BAT formulates tracking as a block assignment process, where blocks with the same label form the appearance of one object and the main advantage of BAT is that it can track an object even when all the part detectors fail as long as the object has assigned blocks.
- 5
- Mixture of Global and Local Experts With Diffusion Transformer for Controllable Face Generation4
Face-MoGLE significantly outperforms state-of-the-art controllable face generation methods in both unimodal and multimodal settings, achieving superior image quality, semantic alignment, aesthetic preference, and zero-shot generalization.
- 4
- 4
- Automatically designing counterfactual regret minimization algorithms for solving imperfect-information games4
This work proposes AutoCFR, a systematic framework that meta-learns novel CFR algorithms through evolution, easing the burden of manual algorithm design and establishing it as a general-purpose framework for solving imperfect-information games.
- Diverse Policies Recovering via Pointwise Mutual Information Weighted Imitation Learning4
This paper enhances the vanilla behavioral cloning learning objective after inferring or assigning a latent style for a trajectory, and incorporates a weighting mechanism based on pointwise mutual information that reflects the significance of each state-action pair's contribution to learning the style.
- PanBench: Towards High-Resolution and High-Performance Pansharpening4
PanBench is introduced, a high-resolution multi-scene dataset containing all mainstream satellites and comprising 5,898 pairs of samples to achieve high-fidelity synthesis of pansharpened images and a Cascaded Multiscale Fusion Network (CMFNet) for Pansharpening is proposed.
- 4
- Anchor-Free One-Stage Online Multi-object Tracking4
An anchor-free one-stage learning framework to perform target detection and appearance embedding in a unified network, which learns for each point in the feature pyramid of the input image an object detection prediction and a feature representation, and an improved non-maximum suppression procedure is introduced.
- Learning Deep Decentralized Policy Network by Collective Rewards for Real-Time Combat Game4
To train DDPN effectively, a novel two-stage learning algorithm is proposed which combines imitation learning from opponent and reinforcement learning by no-regret dynamics and significantly outperforms many state-of-the-art approaches.
- 4
- 4
- 4
- 4
- 4
- Goal-Oriented Skill Abstraction for Offline Multi-Task Reinforcement Learning3
This work proposes Goal-Oriented Skill Abstraction (GO-Skill), a novel approach designed to extract and utilize reusable skills to enhance knowledge transfer and task performance and introduces a skill enhancement phase to refine the extracted skills.
- Position-Aware Guided Point Cloud Completion with CLIP Model3
This work proposes a rapid and efficient method to expand an unimodal framework into a multimodal framework that incorporates a position-aware module designed to enhance the spatial information of the missing parts through a weighted map learning mechanism and outperforms state-of-the-art point cloud completion methods.
- Exploring Reliable PPG Authentication on Smartwatches in Daily Scenarios3
The proposed MTL-RAPID is an efficient and reliable PPG authentication model that employs a multitask joint training strategy, simultaneously assessing signal quality and verifying user identity, and results in a structure that outperforms models trained on individual tasks separately, achieving stronger performance with fewer parameters.
- 3
- 3
- Stabilizing video facial landmark detection and tracking via global and local filtering3
The main contribution within the solution is a novel global and local filtering strategy, which guarantees the robustness of global whole facial shape tracking and the adaptivity of local facial parts tracking.
- Realtime Human Segmentation in Video3
A new real-time video-based human segmentation framework which is designed for the single person from videos to produces smoothing, accurate and fast human segmentsation results.
- ZoomNet: Deep Aggregation Learning for High-Performance Small Pedestrian Detection3
A deep architecture is presented, refereed to as ZoomNet, which performs small pedestrian detection by deep aggregation learning without up-sampling the input and learns and aggregates deep feature representations at multiple levels and retains the spatial information of the pedestrian from different scales.
- 3
- Dual-Pathway Diffusion for Hand Correction in Synthetic Portraits: Global Context Aware and Local Structure Refinement2
A novel dual-pathway diffusion malformed hand correction method, named DD-MHC, which integrates a Global Context Aware Module (GCAM) and a Local Structure Refinement Module (LSRM) through a dual-path architecture, and significantly outperforms existing competing methods in enhancing the semantic consistency between the corrected hand region and the overall image.
- Game-Theoretic Co-Evolution for LLM-Based Heuristic Discovery2
Algorithm Space Response Oracles (ASRO), a game-theoretic framework that reframes heuristic discovery as a program level co-evolution between solver and instance generator, achieving substantially improved generalization and robustness on diverse and out-of-distribution instances.
- K^2-Agent: Co-Evolving Know-What and Know-How for Hierarchical Mobile Device Control2
K2-Agent is a hierarchical framework that models human-like cognition by separating and co-evolving declarative and procedural knowledge for planning and execution by separating and co-evolving declarative and procedural knowledge for planning and execution.
- Deep (Predictive) Discounted Counterfactual Regret Minimization2
An efficient model-free neural CFR algorithm is proposed, overcoming the limitations of existing methods in approximating advanced CFR variants and exhibits faster convergence in typical imperfect-information games and demonstrates stronger adversarial performance in a large poker game.
- DeCoDe: Defer-and-Complement Decision-Making via Decoupled Concept Bottleneck Models2
DeCoDe makes strategy decisions based on human-interpretable concept representations, enhancing transparency throughout the decision process, and significantly outperforms AI-only, human-only, and traditional deferral baselines, while maintaining strong robustness and interpretability even under noisy expert annotations.
- Synergizing Reinforcement Learning and Genetic Algorithms for Neural Combinatorial Optimization2
This work proposes the Evolutionary Augmentation Mechanism (EAM), a general and plug-and-play framework that synergizes the learning efficiency of DRL with the global search power of GAs and demonstrates that EAM significantly improves both solution quality and training efficiency over competitive baselines.
- GTR-Turbo: Merged Checkpoint is Secretly a Free Teacher for Agentic VLM Training2
GTR-Turbo is introduced, a highly efficient upgrade to GTR that matches its performance without training on or querying an expensive teacher model and removes dependence on privileged VLMs, mitigates the"entropy collapse" observed in prior work, and maintains stable training.
- 2
- AdaMemento: Adaptive Memory-Assisted Policy Optimization for Reinforcement Learning2
AdaMemento, an adaptive memory-enhanced RL framework that exploits both positive and negative experiences by learning to predict known local optimal policies based on real-time states, is proposed and theoretically proves the superiority of the new intrinsic motivation and ensemble mechanism.
- OpenHoldem: A Benchmark for Large-Scale Imperfect-Information Game Research2
OpenHoldem is presented, an integrated benchmark for large-scale imperfect-information game research using NLTH and makes three main contributions to this research direction: a standardized evaluation protocol for thoroughly evaluating different NLTH AIs, and an online testing platform with easy-to-use APIs for public NLTH AI evaluation.
- 2
- 2
- 2
- 2
- 2
- MFAD: A Multi-modality Face Anti-spoofing Dataset2
The influences of different modalities on the performance of face anti-spoofing are studied by extensively experimenting with different combinations of modalities with an end-to-end deep learning model, and it is found that the HDR information contributes most among differentmodalities.
- 2
- Non-negative Sparse Coding Using Independent Multi-Codebooks for Near-Duplicate Image Detection2
An efficient approach for detecting near-duplicate images and adopting non-negative sparse coding (NSC) technique to encode features and designing an improved intersection kernel (IIK) to compute image similarity is proposed.
- Scene transformation for detector adaptation2
A novel approach for detection adaptation based on scene transformation, which contributes in both view transformation and automatic parameter estimation, which achieves significant improvements over the pre-trained detector.
- 2
- Hierarchical fusion of local and global visual features with mixture-of-experts for remote sensing image scene classification1
A parallel heterogeneous encoder designed to achieve effective local-global co-representation, which progressively aggregates multi-scale features from both pathways, enabling dynamic cross-level feature interaction and contextual reconstruction to produce highly discriminative representations.
- 1
- DrawMotion: Generating 3D Human Motions by Freehand Drawing1
To accurately capture users’ intended motions without requiring tedious textual input, an algorithm to automatically generate hand-drawn stickman sketches across different dataset formats is developed, and a 2D trajectory condition is incorporated into DrawMotion to achieve improved global spatial control.
- Dual-Flow: Transferable Multi-Target, Instance-Agnostic Attacks via In-the-wild Cascading Flow Optimization1
This work proposes a novel Dual-Flow framework for multi-target instance-agnostic adversarial attacks, utilizing Cascading Distribution Shift Training to develop an adversarial velocity function, and shows substantially stronger robustness against defense mechanisms, such as adversarially trained models.
- 1
- ARFA: An Asymmetric Receptive Field Autoencoder Model for Spatiotemporal Prediction1
An Asymmetric Receptive Field Autoencoder (ARFA) model is proposed, which introduces corresponding sizes of receptive field modules tailored to the distinct functionalities of the encoder and decoder and demonstrates that ARFA consistently achieves state-of-the-art performance on popular datasets.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- –
- Towards Foresighted AI Cooperators with LLM-driven Decision-Time Planning–
This paper introduces DTPAgent, a novel LLM-driven Decision-Time Planning (DTP) framework, and demonstrates on the popular benchmark, Overcooked, that DTPAgent, built with small-scale LLMs, effectively outperforms various types of baselines when faced with unseen partners.
- –
- Towards Transferable Black-Box Attack via Minimizing Maximum Model Discrepancy–
Minimizing Maximum Model Discrepancy (M3D), a novel framework that crafts adversarial examples with enhanced transferability and robustness against model variations, enables more effective and reliable black-box attack strategies, setting a new benchmark for the field.
- –
- –
- –
- –
- –
- UBG: An Unreal BattleGround Benchmark With Object-Aware Hierarchical Proximal Policy Optimization–
This work developed a 3-D open-world first-person shooter game called Unreal BattleGround (UBG), and proposes the object-aware hierarchically proximal policy optimization (OaH-PPO) method in the UBG, which involves a two-level hierarchy, where the high-level controller is tasked with learning option control, and the low-level workers focus on mastering subtasks.
- –
- –
- –
- The SkatingVerse Workshop & Challenge: Methods and Results–
A brief summary of the SkatingVerse Workshop&Challenge is provided including brief introductions to the top three methods including brief introductions to the top three methods.
- –
- –
- Shuffled Autoregression for Motion Interpolation–
This work proposes a novel framework, referred to as Shuffled AutoRegression, which expands the autoregression to generate in arbitrary (shuffled) order and models any inter-frame dependencies as a directed acyclic graph, with an approach to constructing a particular kind of dependency graph.
- Game Interactive Learning: A New Paradigm towards Intelligent Decision-Making–
This work proposes game interactive learning, a novel paradigm as a new approach towards intelligent decision-making in complex and adversarial environments, and performs preliminary experimental verification of the proposed paradigm on two challenging decision-making tasks in tactical-level War-game scenarios.
- –
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.