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
- Suggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works540 from public data
- Point-BERT: Pre-training 3D Point Cloud Transformers with Masked Point Modeling1,232
Point-BERT, a new paradigm for learning Transformers to generalize the concept of BERT to 3D point cloud, is presented and it is shown that a pure Transformer architecture attains 93.8% accuracy on ModelNet40 and 83.1% accuracy in the hardest setting of ScanObjectNN, surpassing carefully designed point cloud models with much fewer hand-made designs.
- DenseCLIP: Language-Guided Dense Prediction with Context-Aware Prompting886
This work presents a new framework for dense prediction by implicitly and explicitly leveraging the pre-trained knowledge from CLIP and converts the original image-text matching problem in CLIP to a pixel- text matching problem and uses the pixel-text score maps to guide the learning of dense prediction models.
- 740
- PoinTr: Diverse Point Cloud Completion with Geometry-Aware Transformers679
A new method is presented that reformulates point cloud completion as a set-to-set translation problem and design a new model, called PoinTr, that adopts a transformer encoder-decoder architecture for point clouds completion that outperforms state-of-the-art methods by a large margin.
- Tri-Perspective View for Vision-Based 3D Semantic Occupancy Prediction598
A tri-perspective view (TPV) representation which accompanies BEV with two additional perpendicular planes is proposed, and it is demonstrated for the first time that using only camera inputs can achieve comparable performance with LiDAR-based methods on theLiDAR segmentation task on nuScenes.
- A Siamese Long Short-Term Memory Architecture for Human Re-identification596
A novel siamese Long Short-Term Memory (LSTM) architecture that can process image regions sequentially and enhance the discriminative capability of local feature representation by leveraging contextual information.
- 551
- Runtime Neural Pruning536
A Runtime Neural Pruning (RNP) framework which prunes the deep neural network dynamically at the runtime and preserves the full ability of the original network and conducts pruning according to the input image and current feature maps adaptively.
- LRRNet: A Novel Representation Learning Guided Fusion Network for Infrared and Visible Images493
A learnable representation approach to the fusion task is adopted, in which the construction of the fusion network architecture is guided by the optimisation algorithm producing the learnable model, and an end-to-end lightweight fusion network is constructed to fuse infrared and visible light images.
- COIN: A Large-Scale Dataset for Comprehensive Instructional Video Analysis477
A simple yet effective method to capture the dependencies among different steps, which can be easily plugged into conventional proposal-based action detection methods for localizing important steps in instruction videos is proposed.
- 472
- SurroundOcc: Multi-Camera 3D Occupancy Prediction for Autonomous Driving459
A SurroundOcc method to predict the 3D occupancy with multi-camera images, which fuse multi-frame LiDAR scans of dynamic objects and static scenes separately and adopts Poisson Reconstruction to fill the holes and voxelize the mesh to get dense occupancy labels.
- 450
- 444
- Stochastic Trajectory Prediction via Motion Indeterminacy Diffusion438
This paper presents a new framework to formulate the trajectory prediction task as a reverse process of motion indeterminacy diffusion (MID), in which the history behavior information and the social interactions are encoded as a state embedding and a Transformer-based diffusion model is devised to capture the temporal dependencies of trajectories.
- Structure-Preserving Super Resolution With Gradient Guidance414
A structure-preserving super resolution method which exploits gradient maps of images to guide the recovery in two aspects and proposes a gradient loss which imposes a second-order restriction on the super-resolved images.
- 399
- Counterfactual Attention Learning for Fine-Grained Visual Categorization and Re-identification378
This paper analyzes the effect of the learned visual attention on network prediction through counterfactual intervention and maximize the effect to encourage the network to learn more useful attention for fine-grained image recognition.
- WebFace260M: A Benchmark Unveiling the Power of Million-Scale Deep Face Recognition375
A distributed framework is developed to train face recognition models efficiently without tampering with the performance, and comprehensive baselines are established on a rich-attribute test set under FRUITS-100ms/500ms/1000ms protocol.
- Unleashing Text-to-Image Diffusion Models for Visual Perception366
It is shown that vision-language pre-trained diffusion models can be faster adapted to downstream visual perception tasks using the proposed VPD, a new framework that exploits the semantic information of a pre-trained text-to-image diffusion model in visual perception tasks.
- 363
- Objects are Different: Flexible Monocular 3D Object Detection349
A flexible framework for monocular 3D object detection which explicitly decouples the truncated objects and adaptively combines multiple approaches for object depth estimation, which outperforms the state-of-the-art method.
- Discriminative multi-manifold analysis for face recognition from a single training sample per person349
- OpenOccupancy: A Large Scale Benchmark for Surrounding Semantic Occupancy Perception321
In the OpenOccupancy benchmark, the large-scale nuScenes dataset is extended with dense semantic occupancy annotations and the Cascade Occupancy Network (CONet) is proposed to refine the coarse prediction, which relatively enhances the performance by ∼30% than the baseline.
- 306
- DensePoint: Learning Densely Contextual Representation for Efficient Point Cloud Processing300
DensePoint is proposed, a general architecture to learn densely contextual representation for point cloud processing that extends regular grid CNN to irregular point configuration by generalizing a convolution operator, which holds the permutation invariance of points, and achieves efficient inductive learning of local patterns.
- NerfingMVS: Guided Optimization of Neural Radiance Fields for Indoor Multi-view Stereo293
A new multi-view depth estimation method that utilizes both conventional SfM reconstruction and learning-based priors over the recently proposed neural radiance fields (NeRF), with surprising findings presented on the effectiveness of correspondence-based opti-mization and NeRF-based optimization over the adapted depth priors.
- BEVerse: Unified Perception and Prediction in Birds-Eye-View for Vision-Centric Autonomous Driving291
The multi-task BEVerse outperforms existing single-task methods on 3D object detection, semantic map construction, and motion prediction and favors in significantly improved efficiency.
- 279
- 278
- 275
- 222
- 218
- 214
- 213
- 205
- FineDiving: A Fine-grained Dataset for Procedure-aware Action Quality Assessment203
This work proposes a procedure-aware approach for action quality assessment, learned by a new Temporal Segmentation Attention module, and proposes to parse pairwise query and exemplar action instances into consecutive steps with diverse semantic and temporal correspondences.
- Uncertainty-Aware Score Distribution Learning for Action Quality Assessment202
An uncertainty-aware score distribution learning (USDL) approach for action quality assessment (AQA) is proposed, which regards an action as an instance associated with a score distribution, which describes the probability of different evaluated scores.
- 202
- Deep Face Super-Resolution With Iterative Collaboration Between Attentive Recovery and Landmark Estimation201
Results show the proposed FSR method with iterative collaboration between two recurrent networks which focus on facial image recovery and landmark estimation significantly outperforms state-of-the-art FSR methods in recovering high-quality face images.
- 197
- AdaPoinTr: Diverse Point Cloud Completion With Adaptive Geometry-Aware Transformers190
A Transformer encoder-decoder architecture, called PoinTr, is proposed, which reformulates point cloud completion as a set-to-set translation problem and employs a geometry-aware block to model local geometric relationships explicitly and can achieve promising performance with higher throughputs and fewer FLOPs compared with the previous best methods in practice.
- DiffTalk: Crafting Diffusion Models for Generalized Audio-Driven Portraits Animation190
This paper turns attention to the emerging powerful Latent Diffusion Models, and model the Talking head generation as an audio-driven temporally coherent denoising process (DiffTalk), and investigates the control mechanism of the talking face, and incorporates reference face images and landmarks as conditions for personality-aware generalized synthesis.
- Oryx MLLM: On-Demand Spatial-Temporal Understanding at Arbitrary Resolution188
Oryx offers an on-demand solution to seamlessly and efficiently process visual inputs with arbitrary spatial sizes and temporal lengths through two core innovations: a pre-trained OryxViT model that can encode images at any resolution into LLM-friendly visual representations and a dynamic compressor module that supports 1x to 16x compression on visual tokens by request.
- Hardness-Aware Deep Metric Learning187
This paper performs linear interpolation on embeddings to adaptively manipulate their hard levels and generate corresponding label-preserving synthetics for recycled training, so that information buried in all samples can be fully exploited and the metric is always challenged with proper difficulty.
- 182
- 181
- SelfOcc: Self-Supervised Vision-Based 3D Occupancy Prediction178
SelfOcc is proposed to explore a self-supervised way to learn 3D occupancy using only video sequences and produces high-quality depth and achieves state-of-the-art results on novel depth synthesis, monocular depth estimation, and surround-view depth estimation on the SemanticKITTI, KITTI-2015, and nuScenes, respectively.
- Group-aware Contrastive Regression for Action Quality Assessment176
It is shown that the relations among videos can provide important clues for more accurate action quality assessment during both training and inference, and a new Contrastive Regression framework is proposed to learn the relative scores by pair-wise comparison, which highlights the differences between videos and guides the models to learning the key hints for assessment.
- 174
- Large Margin Multi-metric Learning for Face and Kinship Verification in the Wild174
A new large margin multi-metric learning (LM\(^3\)L) method for face and kinship verification in the wild that jointly learns multiple distance metrics under which the correlations of different feature representations of each sample are maximized.
- 173
- Streaming 4D Visual Geometry Transformer171
This work proposes a streaming visual geometry transformer that shares a similar philosophy with autoregressive large language models and uses temporal causal attention and cache the historical keys and values as implicit memory to enable efficient streaming long-term 3D reconstruction.
- Diffusion-SDF: Text-to-Shape via Voxelized Diffusion169
Experimental results show that Diffusion-SDF generates both higher quality and more diversified 3D shapes that conform well to given text descriptions when compared to previous approaches.
- Deep Fitting Degree Scoring Network for Monocular 3D Object Detection168
A deep fitting degree scoring network for monocular 3D object detection, which aims to score fitting degree between proposals and object conclusively, and proposes FQNet, which can infer the 3D IoU between the3D proposals and the object solely based on 2D cues.
- 168
- Graininess-Aware Deep Feature Learning for Pedestrian Detection161
This paper proposes a graininess-aware deep feature learning method for pedestrian detection which encodes fine-grained attention masks into convolutional feature maps, which significantly suppresses background interference and highlights pedestrians.
- 160
- Gait recognition for human identification based on ICA and fuzzy SVM through multiple views fusion159
One multiple views fusion recognition approach on the decision level based on product of sum (POS) rule is introduced to overcome the limitation of most single view recognition methods, which achieves better performance than the traditional rank-based fusion rules.
- 158
- 158
- 157
- 157
- 155
- 153
- Kinship verification from facial images under uncontrolled conditions153
This paper collects a kinship face database by searching 400+ pairs of public figures and celebrities from the internet, and automatically detect them with the Viola-Jones face detector and proposes a new spatial pyramid learning-based (SPLE) feature descriptor for face representation and applies support vector machine (SVM) for kinship verification.
- Global-Local Bidirectional Reasoning for Unsupervised Representation Learning of 3D Point Clouds152
This work hypothesizes that a powerful representation of a 3D object should model the attributes that are shared between parts and the whole object, and distinguishable from other objects, and proposes to learn point cloud representation by bidirectional reasoning between the local structures at different abstraction hierarchies and the global shape without human supervision.
- 140
- Scene recognition with objectness139
Experimental results on three widely used scene recognition datasets show that the feature description method called semantic descriptor with objectness (SDO) outperforms the state-of-the-art methods.
- Embodied Task Planning with Large Language Models137
Experimental results show that the generated plan from the TaPA framework can achieve higher success rate than LLaVA and GPT-3.5 by a sizable margin, which indicates the practicality of embodied task planning in general and complex environments.
- 133
- 130
- BridgeNet: A Continuity-Aware Probabilistic Network for Age Estimation128
The proposed BridgeNet, which aims to mine the continuous relation between age labels effectively, outperforms the state-of-the-art methods and can be jointly learned in an end-to-end way.
- PV-RAFT: Point-Voxel Correlation Fields for Scene Flow Estimation of Point Clouds127
Experimental results show that PV-RAFT outperforms state-of-the-art methods by remarkable margins and is evaluated on the FlyingThings3D and KITTI Scene Flow 2015 datasets.
- 127
- 127
- Structure-Preserving Image Super-Resolution123
A structure-preserving super-resolution (SPSR) method to alleviate the above issue while maintaining the merits of GAN-based methods to generate perceptual-pleasant details, and proposes two self-supervised structure learning methods, contrastive prediction and solving jigsaw puzzles, to train the NSEs.
- Gabor-based gradient orientation pyramid for kinship verification under uncontrolled environments123
When combined with the discriminative support vector machine (SVM) classifier, GGOP demonstrates the best performance in the authors' experiments, in comparison with several state-of-the-art face feature descriptors.
- 121
- 120
- Diverse Sample Generation: Pushing the Limit of Generative Data-Free Quantization117
This paper presents a generic Diverse Sample Generation (DSG) scheme for the generative data-free quantization, to mitigate detrimental homogenization and strengthens the loss impact of the specific BN layers for different samples and inhibit the correlation among samples in the generation process.
- Deep Variational Metric Learning117
This paper proposes a deep variational metric learning (DVML) framework to explicitly model the intra-class variance and disentangle the intra -class invariance, namely, the class centers, and can simultaneously generate discriminative samples to improve robustness.
- 3DCNN-DQN-RNN: A Deep Reinforcement Learning Framework for Semantic Parsing of Large-Scale 3D Point Clouds117
A 3DCNN-DQN-RNN method which fuses the 3D convolutional neural network, Deep Q-Network and Residual recurrent neural network for an efficient semantic parsing of large-scale 3D point clouds and provides an automatic process that maps the raw data to the classification results.
- 117
- UniGoal: Towards Universal Zero-shot Goal-oriented Navigation116
The UniGoal achieves state-of-the-art zero-shot performance on three studied navigation tasks with a single model, even outperforming task-specific zero-shot methods and supervised universal methods.
- 115
- 113
- SurroundDepth: Entangling Surrounding Views for Self-Supervised Multi-Camera Depth Estimation112
This paper proposes a SurroundDepth method to incorporate the information from multiple surrounding views to predict depth maps across cameras, and adopts the two-frame structure-from-motion to extract scale-aware pseudo depths to pretrain the models.
- 112
- 111
- 108
- 108
- Bridge-Prompt: Towards Ordinal Action Understanding in Instructional Videos107
This paper reformulates the individual action labels as integrated text prompts for super-vision, which bridge the gap between individual action semantics, and proposes a prompt-based framework, Bridge-Prompt, to model the semantics across adjacent actions, so that it simultaneously exploits both out-of-context and contextual information from a series of ordinal actions in instructional videos.
- 105
- 105
- Uncorrelated discriminant simplex analysis for view-invariant gait signal computing104
A new supervised manifold learning algorithm, called uncorrelated discriminant simplex analysis (UDSA), is proposed for view-invariant gait signal computing, to seek a mapping to project human gait sequences collected from different views into a low-dimensional feature subspace, such that intraclass geometrical structures are preserved and interclass distances of gait sequence are maximized simultaneously.
- 103
- 102
- 102
- Deep Credible Metric Learning for Unsupervised Domain Adaptation Person Re-identification100
Experimental results demonstrate that the proposed deep credible metric learning (DCM-L) method explores credible and valuable training data and improves the performance of unsupervised domain adaptation.
- 99
- WorldDreamer: Towards General World Models for Video Generation via Predicting Masked Tokens98
This work introduces WorldDreamer, a pioneering world model to foster a comprehensive comprehension of general world physics and motions, which significantly enhances the capabilities of video generation and showcases versatility in executing tasks such as text-to-video conversion, image-tovideo synthesis, and video editing.
- Human Trajectory Prediction via Counterfactual Analysis98
This work proposes a counterfactual analysis method for human trajectory prediction to investigate the causality between the predicted trajectories and input clues and alleviate the negative effects brought by the environment bias.
- 98
- 96
- 94
- 94
- OccSora: 4D Occupancy Generation Models as World Simulators for Autonomous Driving91
A diffusion-based 4D occupancy generation model, OccSora, is proposed to simulate 3D world evolution for autonomous driving and has the potential to serve as a world simulator for autonomous driving decision-making.
- 91
- Learning Invariant Color Features for Person Reidentification90
This paper proposes an approach for learning color patterns from pixels sampled from images across two camera views using color as the only cue, and compares it with all the photometric invariant color spaces and shows superior performance over all of them.
- 89
- Collaborative Deep Reinforcement Learning for Multi-object Tracking89
A deep prediction-decision network is developed in the C-DRL, which simultaneously detects and predicts objects under a unified network via deep reinforcement learning.
- Latent Diffusion Model without Variational Autoencoder87
SVG is introduced, a novel latent diffusion model without variational autoencoders, which leverages self-supervised representations for visual generation and preserves the semantic and discriminative capabilities of the underlying self-supervised representations, providing a principled pathway toward task-general, high-quality visual representations.
- 86
- 86
- SOSD-Net: Joint semantic object segmentation and depth estimation from monocular images83
This paper introduces the concept of semantic objectness to exploit the geometric relationship of these two tasks through an analysis of the imaging process, and proposes a Semantic Object Segmentation and Depth Estimation Network (SOSD-Net) based on the objectness assumption, the first network that exploits the geometry constraint for simultaneous monocular depth estimation and semantic segmentation.
- Temporal Coherence or Temporal Motion: Which Is More Critical for Video-Based Person Re-identification?83
This paper proposes a simple yet effective Adversarial Feature Augmentation (AFA) method, which highlights the temporal coherence features by introducing adversarial augmented temporal motion noise.
- 83
- 81
- BiDet: An Efficient Binarized Object Detector80
This paper proposes a binarized neural network learning method called BiDet that generalizes the information bottleneck (IB) principle to object detection, where the amount of information in the high-level feature maps is constrained and the mutual information between the feature maps and object detection is maximized.
- Learning Discriminative Aggregation Network for Video-Based Face Recognition and Person Re-identification79
Experimental results show that the proposed discriminative aggregation network method can generate discriminate images from video clips and improve the overall recognition performance in both the speed and the accuracy for video-based face recognition and person re-identification.
- 78
- SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation76
It is discovered that the locations of instances matter for both instance and semantic 3D scene segmentation, and a weakly-supervised point cloud segmentation method is designed that only requires clicking on one point per instance to indicate its location for annotation.
- 76
- 75
- 75
- 75
- PointOcc: Cylindrical Tri-Perspective View for Point-based 3D Semantic Occupancy Prediction73
A cylindrical tri-perspective view to represent point clouds effectively and comprehensively and a PointOcc model to process them efficiently are proposed and demonstrated that the proposed PointOcc achieves state-of-the-art performance with much faster speed.
- 72
- 72
- Personalized Trajectory Prediction via Distribution Discrimination71
A distribution discrimination method to predict personalized motion patterns by distinguishing the potential distributions, which can be integrated with existing multi-modal stochastic predictive models as a plug-and-play module to learn the more discriminative latent distribution.
- Structured general and specific multi-view subspace clustering71
An efficient and effective optimization approach to solve the objective function of which the time and convergence analyses are provided and the proposed framework can be applied into many existing multi-view subspace clustering methods.
- 71
- 70
- 69
- 69
- 68
- Chain-of-Spot: Interactive Reasoning Improves Large Vision-Language Models67
This work introduces the Chain-of-Spot (CoS) method, which is described as Interactive Reasoning, a novel approach that enhances feature extraction by focusing on key regions of interest (ROI) within the image, corresponding to the posed questions or instructions, thereby offering multi-granularity image features.
- LOGO: A Long-Form Video Dataset for Group Action Quality Assessment67
This work designs a group-aware attention module, which can be easily plugged into existing AQA methods, to enrich the clip-wise representations based on contextual group information and achieves state-of-the-art on the LOGO dataset.
- 66
- 65
- Dynamic Spatial Sparsification for Efficient Vision Transformers and Convolutional Neural Networks64
A dynamic token sparsification framework to prune redundant tokens progressively and dynamically based on the input to accelerate vision Transformers and finds that the idea of adaptive and asymmetric computation can be a general solution for accelerating various architectures.
- Spike Transformer: Monocular Depth Estimation for Spiking Camera64
This paper presents Spike Transformer (Spike-T), a novel paradigm for learning spike data and estimating monocular depth from continuous spike streams, and is the first time that directly depth estimation from spike streams becomes possible.
- EmbodiedSAM: Online Segment Any 3D Thing in Real Time63
A geometric-aware query lifting module is proposed to represent the 2D masks generated by SAM by 3D-aware queries, which are transferred to fine-grained shapes on 3D point clouds and benefit from the query representation for 3D masks, which enables real-time inference.
- 63
- 63
- Shapley-NAS: Discovering Operation Contribution for Neural Architecture Search62
A Shapley value based method to evaluate operation contribution (Shapley-NAS) for neural architecture search that outperforms the state-of-the-art methods by a considerable margin with light search cost.
- 62
- Deep Reinforcement Learning with Iterative Shift for Visual Tracking62
A deep reinforcement learning with iterative shift (DRL-IS) method for single object tracking, where an actor-critic network is introduced to predict the iterative shifts of object bounding boxes, and evaluate the shifts to take actions on whether to update object models or re-initialize tracking.
- 61
- 60
- RandomRooms: Unsupervised Pre-training from Synthetic Shapes and Randomized Layouts for 3D Object Detection60
A new method called RandomRooms is proposed to generate random layouts of a scene by making use of the objects in the synthetic CAD dataset and learn the 3D scene representation by applying object-level contrastive learning on two random scenes generated from the same set of synthetic objects.
- Summarizing surveillance videos with local-patch-learning-based abnormality detection, blob sequence optimization, and type-based synopsis60
An abnormality-type-based method which creates short-period summary videos from long-period input surveillance videos by properly arranging abnormal blob sequences according to their activity types is proposed.
- 59
- 59
- Towards Interpretable Deep Metric Learning with Structural Matching58
This paper proposes a structural matching strategy that explicitly aligns the spatial embeddings by computing an optimal matching flow between feature maps of the two images and enables deep models to learn metrics in a more human-friendly way.
- Multi-modal uniform deep learning for RGB-D person re-identification58
A multi-modal fusion layer is designed to combine features extracted from both depth images and RGB images through the network with a uniform latent variable which is robust to noise, and optimize the fusion layer with two CNN networks jointly.
- Cascade Subspace Clustering58
This work proposes a deep learning based subspace clustering method which simultaneously learns a compact representation using a neural network and a clustering assignment by minimizing the discrepancy between pair-wise sample-centers distributions.
- 57
- 57
- 57
- Multi-modal Unsupervised Feature Learning for RGB-D Scene Labeling57
This paper adapts the unsupervised feature learning technique for RGB-D labeling as a multi-modality learning problem and performs feature learning and feature encoding simultaneously which significantly boosts the performance.
- OccNeRF: Advancing 3D Occupancy Prediction in LiDAR-Free Environments56
This paper parameterize the reconstructed occupancy fields and reorganize the sampling strategy to align with the cameras’ infinite perceptive range, and proposes an OccNeRF method for training occupancy networks without 3D ground truth.
- 56
- 56
- Mining Semantics-Preserving Attention for Group Activity Recognition56
A Semantics-Preserving Teacher-Student (SPTS) model for group activity recognition in videos, which aims to mine the semantics-preserving attention to automatically seek the key people and discard the misleading people.
- Towards Accurate Post-Training Quantization for Diffusion Models55
An accurate post-training quantization framework of diffusion models (APQ-DM) for efficient image generation by designing distribution-aware quantization functions for activation discretization in different timesteps and searching the optimal timesteps for informative calibration image generation to reduce the discretization errors with negligible computational overhead.
- 55
- 55
- Multi-feature multi-manifold learning for single-sample face recognition54
Experimental results on the widely used AR, FERET and LFW datasets demonstrate the efficacy of the proposed M^3L method, and a multi-feature manifold-manifold distance measure is presented to recognize the probe subjects.
- Doe-1: Closed-Loop Autonomous Driving with Large World Model53
A closed-loop framework for autonomous driving and a large Driving wOrld modEl (Doe-1) for unified perception, prediction, and planning is explored and a multi-modal transformer is trained to autoregressively generate perception, prediction, and planning tokens in an end-to-end and unified manner.
- FGR: Frustum-Aware Geometric Reasoning for Weakly Supervised 3D Vehicle Detection53
A frustum-aware geometric reasoning (FGR) method to detect vehicles in point clouds without any 3D annotations and reaches comparable performance with fully supervised methods on the KITTI dataset.
- Nonlinear subspace clustering for image clustering53
A nonlinear subspace clustering method for image clustering that employs a feed-forward neural network to map samples into a nonlinear space and performs subspace clusters at the top layer of the network so that the mapping functions and the clustering issues are iteratively learned.
- 53
- Graph-Based Social Relation Reasoning52
A simpler, faster, and more accurate method named graph relational reasoning network (GR2N) for social relation recognition, which considers the paradigm of jointly inferring the relations by constructing a social relation graph.
- 52
- 52
- 51
- 51
- 51
- 51
- 50
- 50
- Part-Activated Deep Reinforcement Learning for Action Prediction50
This paper designs the PA-DRL to exploit the structure of the human body by extracting skeleton proposals under a deep reinforcement learning framework and considers the saliency part for expressing actions.
- 50
- EmbodiedOcc: Embodied 3D Occupancy Prediction for Vision-Based Online Scene Understanding49
This paper formulate an embodied 3D occupancy prediction task to target this practical scenario and proposes a Gaussian-based EmbodiedOcc framework to accomplish it, which outperforms existing methods by a large margin and accomplishes the embodied occupancy prediction with high accuracy and efficiency.
- 49
- 49
- Self-Calibrated CLIP for Training-Free Open-Vocabulary Segmentation48
Self-Calibrated CLIP (SC-CLIP), a training-free method that calibrates CLIP to generate finer representations while preserving its original generalization ability—without introducing new parameters or relying on additional backbones is proposed.
- 47
- 47
- Rotation-Robust Intersection over Union for 3D Object Detection47
A Rotation-robust Intersection over Union for 3D object detection, which aims to learn the overlap of rotated bounding boxes and is suitable for the detection task of 2D rotated objects, such as text boxes and cluttered targets in the aerial images.
- 47
- 47
- 46
- Structure-Aware Face Clustering on a Large-Scale Graph with 107Nodes46
The proposed STructure-AwaRe Face Clustering (STAR-FC) method is the first to train on very large-scale graph with 20M nodes, and achieve superior inference results on 12M testing data.
- 46
- 45
- Comprehensive Instructional Video Analysis: The COIN Dataset and Performance Evaluation45
This article proposes a large-scale dataset called “COIN” for COmprehensive INstructional video analysis, which contains 11,827 videos of 180 tasks in 12 domains related to the authors' daily life and proposes two simple yet effective methods, which can be easily plugged into conventional proposal-based action detection models.
- MetaDistiller: Network Self-Boosting via Meta-Learned Top-Down Distillation44
This work proposes that better soft targets with higher compatibil-ity can be generated by using a label generator to fuse the feature maps from deeper stages in a top-down manner, and can employ the meta-learning technique to optimize this label generator.
- 43
- 43
- 43
- Take-A-Photo: 3D-to-2D Generative Pre-training of Point Cloud Models42
A novel 3D-to-2D generative pre-training method that is adaptable to any point cloud model and effective in boosting the performance of architecture-oriented approaches, achieving state-of-the-art performance when fine-tuning on ScanObjectNN classification and ShapeNet-Part segmentation tasks.
- Relaxation-Free Deep Hashing via Policy Gradient42
This approach formulates the non-smooth part of the hashing network as sampling with a stochastic policy, so that the retrieval performance degradation caused by the relaxation can be avoided and the differentiation challenge for discrete optimization can be naturally addressed.
- Regularized local metric learning for person re-identification42
This paper proposes a regularized local metric learning (RLML) method for person re-identification which combines global and local metrics to represent the within-class and between-class variances and utilizes the local distribution of the training data to avoid the overfitting problem.
- 42
- Enhanced Gabor-based region covariance matrices for palmprint recognition42
Experimental results demonstrate the efficacy of the enhanced Gabor-based region covariance matrices method for palmprint recognition.
- 41
- 41
- Segment and Caption Anything40
A method to efficiently equip the Segment Anything Model (SAM) with the ability to generate regional captions by introducing a lightweight query-based feature mixer that aligns the region-specific features with the embedding space of language models for later caption generation.
- 40
- HyperDet3D: Learning a Scene-conditioned 3D Object Detector40
This paper proposes a discriminative Multi-head Scene-specific Attention (MSA) module to dynamically control the layer parameters of the detector conditioned on the fusion of scene-conditioned knowledge and achieves state-of-the-art results on the 3D object detection benchmark of the ScanNet and SUN RGB-D datasets.
- Kinship verification in the wild: The first kinship verification competition40
The key goal of this competition is to compare the performance of different methods on a new-collected dataset with the same evaluation protocol and develop the first standardized benchmark for kinship verification in the wild.
- Narrative Action Evaluation with Prompt-Guided Multimodal Interaction39
A prompt-guided multimodal interaction framework that outperforms separate learning methods and naive multi-task learning methods in narrative action evaluation and establishes benchmarks for NAE.
- Graph-Based Kinship Reasoning Network39
A graph-based kinship reasoning network for kinship verification, which aims to effectively perform relational reasoning on the extracted features of an image pair, which outperforms the state-of-the-art methods.
- 39
- Learning Series-Parallel Lookup Tables for Efficient Image Super-Resolution38
A framework of series-parallel lookup tables (SPLUT) is proposed to alleviate the above issues and achieve efficient image super-resolution and no longer need interpolation methods which introduce redundant computations so that the method can achieve faster inference speed.
- Generalizable Mixed-Precision Quantization via Attribution Rank Preservation38
This paper observes that locating network attribution correctly is general ability for accurate visual analysis across different data distribution, and preserves attribution rank consistency between the quantized models and their full-precision counterparts via efficient capacity-aware attribution imitation for generalizable mixed- Precision quantization strategy search.
- 38
- FLAG3D: A 3D Fitness Activity Dataset with Language Instruction37
FLAG3D, a large-scale 3D fitness activity dataset with language instruction containing 180K sequences of 60 categories, contributes great research value for various challenges, such as cross-domain human action recognition, dynamic human mesh recovery, and language-guided human action generation.
- 37
- 37
- Introspective Deep Metric Learning36
This paper proposes an introspective deep metric learning (IDML) framework for uncertainty-aware comparisons of images, and proposes an introspective similarity metric to make similarity judgments between images considering both their semantic differences and ambiguities.
- 36
- 36
- 36
- 35
- 35
- 35
- Back to Reality: Weakly-supervised 3D Object Detection with Shape-guided Label Enhancement34
A weakly-supervised approach for 3D object detection, which makes it possible to train a strong 3D detector with position-level annotations and makes use of synthetic 3D shapes to convert the weak labels into fully-annotated virtual scenes as stronger supervision, and utilizes the perfect virtual labels to complement and refine the real labels.
- 34
- 34
- GE-Grasp: Efficient Target-Oriented Grasping in Dense Clutter33
This work presents a generic framework called GE-Grasp for robotic motion planning in dense clutter, where it leverage diverse action primitives for occluded object removal and presents the generator-evaluator architecture to avoid spatial collisions.
- Adaptive Neighborhood Metric Learning33
The proposed novel distance metric learning algorithm, named adaptive neighborhood metric learning (ANML), design two thresholds to adaptively identify the inseparable similar and dissimilar samples in the training procedure, thus inseparable sample removing and metric parameter learning are implemented in the same procedure.
- Conditional Single-View Shape Generation for Multi-View Stereo Reconstruction33
This work introduces a conditional generative network to model the uncertainty for single-view reconstruction, and designs new differentiable guidance including the front constraint, the diversity constraint, and the consistency loss to enable effective single-View conditional generation and multi-view synthesis.
- Meta-Mining Discriminative Samples for Kinship Verification32
A Discriminative Sample Meta-Mining approach to kinship verification, which samples in the unbalanced train batch are re-weighted by the learned meta-miner to optimize the kinship models.
- 32
- 32
- 32
- Efficient nearest neighbor search in high dimensional hamming space31
The proposed method significantly outperforms the state of the arts, obtaining improved search accuracy at various speed up factors, e.g., at least 16% improvement of search accuracy over previous methods (from 67.7% to 83.7%) when the search speed is 200 times faster than the linear scan for a one million database.
- Dual-Agent Deep Reinforcement Learning for Deformable Face Tracking31
This paper proposes a dual-agent deep reinforcement learning (DADRL) method for deformable face tracking, which generates bounding boxes and detects facial landmarks interactively from face videos and proposes a unified framework for simultaneous bounding box tracking and landmark detection.
- Planning Irregular Object Packing via Hierarchical Reinforcement Learning30
Experiments show that the proposed deep hierarchical reinforcement learning approach to simultaneously plan packing sequence and placement for irregular object packing can pack more objects with less time cost than the state-of-the-art packing methods of irregular objects.
- Uncertainty-Aware Representation Learning for Action Segmentation30
An uncertainty-aware representation Learning (UARL) method for action segmentation that designs the UARL to exploit the transitional expression between two action periods by uncertainty learning, which captures the uncertainty of the action and exploits the tendency during the action.
- 30
- 30
- 30
- TCOVIS: Temporally Consistent Online Video Instance Segmentation29
This paper proposes a novel online method for video instance segmentation, called TCOVIS, which fully exploits the temporal information in a video clip, and achieves state-of-the-art performance on all benchmarks without bells-and-whistles.
- 29
- 29
- Reasoning Graph Networks for Kinship Verification: From Star-Shaped to Hierarchical28
This paper investigates the problem of facial kinship verification by learning hierarchical reasoning graph networks by introducing a set of latent reasoning nodes and constructs a hierarchical graph with them and develops a Hierarchical Reasoning Graph Network to exploit more powerful and flexible capacity.
- Latent Fingerprint Registration via Matching Densely Sampled Points28
A non-minutia latent fingerprint registration method which estimates the spatial transformation between a pair of fingerprints through a dense fingerprint patch alignment and matching procedure, which achieves the state-of-the-art registration performance, especially under challenging conditions.
- 28
- Cryptanalysis of reduced versions of the Camellia block cipher28
A flaw in the approach used to choose plaintexts or ciphertexts in certain previously published square-like cryptanalytic results for Camellia is described and two possible approaches to correct them are given.
- Text-guided Sparse Voxel Pruning for Efficient 3D Visual Grounding27
Text-guided pruning and completion-based addition are proposed to deeply fuse 3D scene representation and text features in an efficient way by gradual region pruning and target completion and achieves top inference speed and surpasses previous fastest method by 100% FPS.
- Gait Recognition in the Wild: A Large-Scale Benchmark and NAS-Based Baseline27
The proposed GREW benchmark proves to be essential for both training and evaluating gait recognizers in unconstrained scenarios, and the Single Path One-Shot neural architecture search with uniform sampling for Gait recognition is proposed, named SPOSGait, which is the first NAS-based gait recognition model.
- Reinforced Axial Refinement Network for Monocular 3D Object Detection27
The proposed framework, Reinforced Axial Refinement Network (RAR-Net), serves as a post-processing stage which can be freely integrated into existing monocular 3D detection methods, and improve the performance on the KITTI dataset with small extra computational costs.
- A Vessel-Focused 3D Convolutional Network for Automatic Segmentation and Classification of Coronary Artery Plaques in Cardiac CTA27
Results show that by artery extraction and straightening, the training time is reduced by 40% and the segmentation performance of non-calcified plaques and mixed calcified plaque gains significantly, which shows potential value for clinical application.
- Order-Sensitive Deep Hashing for Multimorbidity Medical Image Retrieval27
This paper designs a deep hash model for powerful feature extraction and preserves the ranking list with a triplet based ranking loss for better assessment assistance and introduces the cross-entropy based multi-label classification loss to exploit multi- label information.
- X-3D: Explicit 3D Structure Modeling for Point Cloud Recognition26
X-3D functions by capturing the explicit local structural information within the input 3D space and employing it to produce dynamic kernels with shared weights for all neighborhood points within the current local region, thereby improving the extraction of local features.
- DPMesh: Exploiting Diffusion Prior for Occluded Human Mesh Recovery26
DPMesh is introduced, an innovative framework for occluded human mesh recovery that capitalizes on the pro-found diffusion prior about object structure and spatial relationships embedded in a pre-trained text-to-image diffusion model to enhance the perception capability for occluded poses.
- 26
- MCUFormer: Deploying Vision Transformers on Microcontrollers with Limited Memory26
A hardware-algorithm co-optimizations method to deploy vision transformers on microcontrollers with extremely limited memory called MCUFormer is proposed, where the one-shot network architecture search (NAS) is generalized to discover the optimal architecture with highest task performance given the memory budget from the micro Controllers.
- 26
- 26
- 26
- 26
- ThinkBot: Embodied Instruction Following with Thought Chain Reasoning25
This work designs an instruction completer based on large language models to recover the missing actions with interacted objects between consecutive human instruction, where the perceived surrounding environments and the completed sub-goals are considered for instruction completion.
- Learning from Temporal Spatial Cubism for Cross-Dataset Skeleton-based Action Recognition25
This article designs two self-supervised learning classification tasks to explore the temporal and spatial dependency of a skeleton-based action and improve the generalization ability of the model.
- 25
- Left atrial appendage segmentation and quantitative assisted diagnosis of atrial fibrillation based on fusion of temporal-spatial information25
The experimental results demonstrate that the approach can construct the 3-D LAA geometries robustly compared to manual annotations, and reasonably infer that the LAA undergoes filling, emptying and re-filling, re-emptying in a cardiac cycle.
- 25
- 25
- 25
- Improved discriminant locality preserving projections for face and palmprint recognition25
It is shown that transforming an image to a diagonal image is equivalent to assigning an appropriate weight to each pixel of the original image to emphasize its different importance for recognition, which provides the rationale and superiority of using diagonal images for 2D subspace learning.
- 25
- SparseMM: Head Sparsity Emerges from Visual Concept Responses in MLLMs24
SparseMM is introduced, a KV-Cache optimization strategy that allocates asymmetric computation budgets to heads in LLMs based on their visual scores, leveraging the sparity of visual heads for accelerating the inference of MLLMs.
- Memory-based Adapters for Online 3D Scene Perception24
An adapter-based plug-and-play module for the backbone of 3D scene perception model, which constructs memory to cache and aggregate the extracted RGB-D features to empower offline models with temporal learning ability, and a queued memory mechanism to cache the supporting point cloud and image features.
- 24
- 23
- 23
- 23
- 23
- 23
- 23
- 23
- Stag-1: Towards Realistic 4D Driving Simulation with Video Generation Model22
A Spatial-Temporal simulAtion for drivinG (Stag-1) model to reconstruct real-world scenes and design a controllable generative network to achieve 4D simulation, which shows promising performance in multi-view scene consistency, background coherence, and accuracy.
- 22
- 22
- Endo3D: Online Workflow Analysis for Endoscopic Surgeries Based on 3D CNN and LSTM22
This paper proposes to extract fine-level temporal features from video clips using 3D convolutional networks (CNN) and use Long Short-Term Memory (LSTM) networks to capture coarse-level information and outperforms state-of-the-art online methods without using specific knowledge of surgeries.
- 22
- 22
- 21
- 21
- Attributable Visual Similarity Learning21
This paper proposes an attributable visual similarity learning (AVSL) framework for a more accurate and ex-plainable similarity measure between images and establishes a bottom-up similarity construction and top-down similarity inference framework to infer the similarity based on semantic hier-archy consistency.
- 21
- 21
- 21
- 21
- 21
- 20
- 20
- 20
- 20
- Structural Deep Metric Learning for Room Layout Estimation20
This paper proposes a structural deep metric learning (SDML) method for room layout estimation, which aims to recover the 3D spatial layout of a cluttered indoor scene from a monocular RGB image.
- 20
- Gait Recognition Using Wavelet Descriptors and Independent Component Analysis20
An approach to automatic gait recognition based on wavelet descriptors and independent component analysis (ICA) for the purpose of human identification at a distance with encouraging recognition accuracy with comparatively low computational cost is proposed.
- 19
- SemAffiNet: Semantic-Affine Transformation for Point Cloud Segmentation19
This work proposes SemAffiNet for point cloud semantic segmentation, which utilizes the attention mechanism in the Transformer module to implicitly and explicitly capture global structural knowledge within local parts for overall comprehension of each category.
- 19
- GC-VLN: Instruction as Graph Constraints for Training-free Vision-and-Language Navigation18
A training-free framework for vision-and-language navigation (VLN) that formulates navigation guidance as graph constraint optimization by decomposing instructions into explicit spatial constraints, enabling zero-shot adaptation to unseen environments.
- 18
- 18
- Skip-Plan: Procedure Planning in Instructional Videos via Condensed Action Space Learning18
The Skip-Plan model explores all sorts of reliable sub-relations within an action sequence in the condensed action space by skipping uncertain nodes and edges in action chains and decomposing relatively long chains into multiple short sub-chains by skipping unreliable intermediate actions.
- 18
- 18
- 18
- 18
- Dominant singular value decomposition representation for face recognition18
A new face representation method to select a subset of important bases and regulate their singular values (SVs) according to their discriminative and reconstructive power simultaneously for face recognition is presented.
- 17
- Procedure-Aware Action Quality Assessment: Datasets and Performance Evaluation17
A new approach is proposed (i.e., spatial-temporal segmentation attention, STSA) that introduces procedure segmentation to parse an action into consecutive steps, learns powerful representations from these steps by constructing spatial motion attention and procedure-aware cross-attention, and designs a fine-grained contrastive regression to achieve an interpretable scoring mechanism.
- MetaAge: Meta-Learning Personalized Age Estimators17
This paper introduces a personalized estimator meta-learner, which takes identity features as the input and outputs the parameters of customized estimators of age estimation, and introduces a meta-learning method named MetaAge for age estimation.
- 17
- 16
- Owl-1: Omni World Model for Consistent Long Video Generation16
An Omni World modeL (Owl-1) is proposed to produce long-term coherent and comprehensive conditions for consistent long video generation and achieves comparable performance with SOTA methods on VBench-I2V and VBench-Long, validating its ability to generate high-quality video observations.
- 16
- 16
- 16
- Similarity-Aware Fusion Network for 3D Semantic Segmentation16
Experimental results on ScanNetV2 benchmark demonstrate that SAFNet outperforms existing state-of-the-art fusion-based approaches across various data integrity.
- 16
- 16
- Left Atrial Appendage Segmentation Based on Ranking 2-D Segmentation Proposals16
This work presents a non-model based semi-automated approach for LAA segmentation on CTA data that requires only manual selection of four fiducial points to obtain the bounding box for the LAA.
- Multi-manifold metric learning for face recognition based on image sets16
A new multi-manifold metric learning method for the task of face recognition based on image sets that aims to learn distance metrics to measure the similarity between manifold pairs and is extensively evaluated on three widely studied face databases.
- Robust gait recognition via discriminative set matching16
A new Multiview Subspace Representation (MSR) method is devised which considers gait sequences collected from different views of the same subject as a feature set and extracts a linear subspace to describe the feature set.
- 16
- Gait Recognition Using Independent Component Analysis16
A simple background subtraction algorithm is introduced to segment the moving figures accurately and to achieve binary silhouettes and a fast and robust fixed-point algorithm named FastICA is adopted for reducing computation cost.
- 15
- 15
- Active appearance models using statistical characteristics of Gabor based texture representation15
A new AAM method based on Gabor texture feature representation is proposed, which is the first time to extract statistical features from both Gabor magnitude and Gabor phase as the texture representation in AAM.
- Gait Recognition via Independent Component Analysis Based on Support Vector Machine and Neural Network15
This paper proposes a method of automatic gait recognition using Fourier descriptors and independent component analysis (ICA) for the purpose of human identification at a distance and results show that this method has encouraging recognition accuracy.
- 14
- SIMPLE: SIngle-network with Mimicking and Point Learning for Bottom-up Human Pose Estimation14
A novel multi-person pose estimation framework, SIngle-network with Mimicking and Point Learning for Bottom-up Human Pose Estimation (SIMPLE), which enables SIMPLE to mimic the pose knowledge from the high-performance top-down pipeline, which significantly promotes SIMPLE's accuracy while maintaining its high efficiency during inference.
- 14
- Regularized Bayesian Metric Learning for Person Re-identification14
Experimental results demonstrate the advantage of the proposed RBML method to model and regulate the eigen-spectrums of these two covariance matrices in a parametric manner, so that discriminating information can be better exploited.
- Content-Aware Warping for View Synthesis13
This paper proposes content-aware warping, which adaptively learns the interpolation weights for pixels of a relatively large neighborhood from their contextual information via a lightweight neural network, and proposes a new end-to-end learning-based framework for novel view synthesis from a set of input source views.
- Learning principal orientations and residual descriptor for action recognition13
The PORD aims to learn the statistic principal orientations and to represent the local features of action videos with residual values and is competitive with the state-of-the-art methods.
- Left Atrium Segmentation in CT Volumes with Fully Convolutional Networks13
An approach to automatically segmenting left atrium in 3D CT volumes using fully convolutional neural networks (FCNs) and achieves Dice coefficient scores over 93% with computation time below 35s per volume, despite of the high variation of LA.
- 13
- 13
- Supervised Discriminative Hashing for Compact Binary Codes13
A family of discriminative hash functions is designed to map data points of the original high-dimensional space into nearby compact binary codes while preserving the geometrical similarity and discriminant properties in both global and local neighborhoods.
- 13
- 13
- Embodied Instruction Following in Unknown Environments12
This work proposes an embodied instruction following (EIF) method for complex tasks in the unknown environment, where the agent efficiently explores the unknown environment to generate feasible plans with existing objects to accomplish abstract instructions.
- 12
- Hardness-Aware Scene Synthesis for Semi-Supervised 3D Object Detection12
A hardness-aware scene synthesis (HASS) method to generate adaptive synthetic scenes to improve the generalization of the detection models and obtain pseudo-labels for unlabeled objects and generate diverse scenes with different compositions of objects and backgrounds.
- Binarizing Sparse Convolutional Networks for Efficient Point Cloud Analysis12
The experimental results show that the proposed BSC-Net achieves significant improvement upon the srtong baseline and outperforms the state-of-the-art network binarization methods by a remarkable margin without additional computation overhead for binarizing sparse convolutional networks.
- 12
- 12
- 12
- Structure-Aware Multimodal Feature Fusion for RGB-D Scene Classification and Beyond12
This article investigates a framework allowing greater spatial flexibility, in which the Fisher vector (FV)-encoded distribution of local CNN features, obtained from a multitude of region proposals per image, is considered instead.
- 12
- 12
- EfficientLLaVA: Generalizable Auto-Pruning for Large Vision-language Models11
This paper formulate the generalization gap of the pruning strategy using the structural risk minimization principle and iteratively search for the optimal pruning policy within a given search space and optimize the vision projector to evolve the search space with higher upper bound of performance.
- 11
- 11
- DSPDet3D: 3D Small Object Detection with Dynamic Spatial Pruning11
This paper theoretically derive a dynamic spatial pruning (DSP) strategy to prune the redundant spatial representation of 3D scene in a cascade manner according to the distribution of objects, and designs and constructs a multi-level 3D detector named DSPDet3D which benefits from high spatial resolution to achieves high accuracy on small object detection, while reducing redundant computation by only focusing on small object areas.
- 11
- Preventing Local Pitfalls in Vector Quantization via Optimal Transport11
OptVQ is introduced, a novel vector quantization method that employs the Sinkhorn algorithm to optimize the optimal transport problem, thereby enhancing the stability and efficiency of the training process and mitigating the influence of diverse data distributions on the Sinkhorn algorithm.
- Dense Hybrid Proposal Modulation for Lane Detection11
This paper presents a dense hybrid proposal modulation method for lane detection that consistently outperform the baseline model on most metrics without introducing new parameters and reducing inference speed.
- AMixer: Adaptive Weight Mixing for Self-attention Free Vision Transformers11
This paper investigates the key differences between vision Transformers and recent all-MLP models and proposes a computationally and parametrically efficient operation named adaptive weight mixing to generate attention weights without token-token interactions and develops a new architecture named AMixer to capture both long-term and short-term spatial dependencies without self-attention.
- Multi-Proxy Wasserstein Classifier for Image Classification11
This work adopts optimal transport theory to calculate a non-uniform matching flow between the elements in the feature map of a sample and the proxies of a class in a closed way, and enables the models to achieve partial matching.
- Spatial Geometric Reasoning for Room Layout Estimation via Deep Reinforcement Learning11
This work model the layout in 3D as a Markov decision process, in which the layout is incrementally adjusted based on the difierence between the current layout and the target image, and the policy is learned via deep reinforcement learning.
- Context Aware 3D Fully Convolutional Networks for Coronary Artery Segmentation11
A context aware 3D fully convolutional network (FCN) for vessel enhancement and segmentation in coronary computed tomography angiography (CTA) volumes is proposed and significantly outperforms conventional Hessian vesselness based approach on a dataset of 50 coronary CTA volumes.
- 11
- 11
- 11
- Path Choice Matters for Clear Attribution in Path Methods10
SAMP is a model-agnostic interpreter, which efficiently searches the near-optimal path from a pre-defined set of manipulation paths and proposes the infinitesimal constraint (IC) and momentum strategy (MS) to improve the rigorousness and optimality.
- Dynamics-Aware Adversarial Attack of Adaptive Neural Networks10
This paper reformulates the gradients to be aware of the potential dynamic changes of network architectures, so that the learned attack better leads the next step than the dynamics-unaware methods when network architecture changes dynamically.
- 10
- 10
- 10
- 10
- 10
- 10
- 10
- 10
- 9
- Temporal Feature Matters: A Framework for Diffusion Model Quantization9
This work introduces a novel quantization framework that includes a novel quantization framework that preserves most of the temporal information and ensures high-quality end-to-end generation of diffusion models.
- 9
- V2M: Visual 2-Dimensional Mamba for Image Representation Learning9
This paper proposes a Visual 2-Dimensional Mamba (V2M) model as a complete solution, which directly processes image tokens in the 2D space and effectively incorporates the 2D locality prior yet inherits the efficiency and input-dependent scalability of Mamba.
- 9
- Estimating 3D Finger Pose via 2D-3D Fingerprint Matching9
A finger specific algorithm for estimating 3D finger pose including roll, pitch, and yaw from fingerprint images, a non-learning algorithm with good generalization ability and robustness in real applications is proposed.
- Instance Similarity Learning for Unsupervised Feature Representation9
This paper employs the Generative Adversarial Networks (GAN) to mine the underlying feature manifold, where the generated features are applied as the proxies to progressively explore the feature manifold so that the semantic similarity among instances is acquired as reliable pseudo supervision.
- 9
- Robust face recognition via gradient-based sparse representation9
By combining image pixels and image gradients, GSRC has less model error and requires fewer training samples from each individual than sparse representation–based classification (SRC), which makes GSRC work much faster than SRC.
- 9
- 9
- 9
- 9
- 8
- Token-Label Alignment for Vision Transformers8
A token-label alignment method to trace the correspondence between transformed tokens and the original tokens to maintain a label for each to-ken, which improves the performance of ViTs on image classification, semantic segmentation, objective detection, and transfer learning tasks.
- Bort: Towards Explainable Neural Networks with Bounded Orthogonal Constraint8
This work introduces Bort, an optimizer for improving model explainability with boundedness and orthogonality constraints on model parameters, derived from the sufficient conditions of model comprehensibility and invertibility.
- Dynamic Metric Learning with Cross-Level Concept Distillation8
This work identifies the negative repulsion as the key obstacle in existing methods since a positive pair is always positive for coarser semantic levels but not for negative pairs, and proposes a hierarchical concept refiner to construct multiple levels of concept embeddings of an image and then pull closer the distance of the corresponding concepts.
- 8
- Deep Hashing with Active Pairwise Supervision8
A Deep Hashing method with Active Pairwise Supervision that achieves the accuracy of supervised hashing methods with only 30% labeled training samples and improves the semi-supervised binary codes by a sizable margin is proposed.
- 8
- 8
- Quantize-then-Rectify: Efficient VQ-VAE Training7
This work demonstrates that a pre-trained VAE can be efficiently transformed into a VQ-VAE by controlling quantization noise within the VAE's tolerance threshold, and demonstrates that ReVQ achieves superior efficiency-reconstruction trade-offs.
- 7
- 7
- Activity-based person identification using sparse coding and discriminative metric learning7
This paper presents a new activity-based person identification method using sparse coding and discriminative metric learning that aims to recognize people from different activities such as running, jumping, skipping, and so on.
- 7
- 7
- 7
- Learning Counterfactually Decoupled Attention for Open-World Model Attribution6
CDAL explicitly models the causal relationships between the attentional visual traces and source model attribution, and counterfactually decouples the discriminative model-specific artifacts from confounding source biases for comparison, so that the resulting causal effect provides a quantification on the quality of learned attention maps.
- 6
- OPERA: Omni-Supervised Representation Learning with Hierarchical Supervisions6
This paper provides a unified perspective of supervisions from labeled and unlabeled data and proposes a unified framework of fully supervised and self-supervised learning.
- 6
- 6
- Facial Kinship Verification6
In this chapter, the background of facial kinship verification is introduced and then the state-of-the-art of Facial Kinship Verification is reviewed, and the organization of the book is outlined.
- 6
- 6
- WhiADD: Semantic-Acoustic Fusion for Robust Audio Deepfake Detection5
A novel audio deepfake detection framework ''WhiADD'' enhanced by semantic-acoustic fusion and cross-modal generalization is proposed, establishing a robust paradigm for combating evolving codec-based forgeries in multilingual and dynamically evolving adversarial scenarios.
- Ambiguity-Aware Point Cloud Segmentation by Adaptive Margin Contrastive Learning5
An adaptive margin contrastive learning method for 3D semantic segmentation on point clouds that leverages predicted ambiguities to enable the ambiguous embeddings to be more reliable, thereby boosting segmentation performance and enhancing robustness.
- 5
- 5
- 5
- 5
- Anyview: Generalizable Indoor 3D Object Detection with Variable Frames5
A geometric learner is proposed to mine the local geometric features of each input RGB-D image frame and implement local-global feature interaction through a designed spatial mixture module, which ensures consistent global feature density and further enhances the generalization after fusion.
- 5
- Guest Editorial Introduction to the Special Section on Representation Learning for Visual Content Understanding5
This special section focuses on data-effective representation learning methods for visual content understanding using the ResNet method for image classification, the DeepFace method for face recognition, and the feature pyramid networks (FPNs) method for object detection.
- Towards Accurate and Complete Registration of Coronary Arteries in CTA Images5
This paper proposes an automated method for accurate and complete registration of coronary arteries that includes bifurcation matching, segment registration, and a novel approach to further improve the completeness of registration by combining the previous results and a level set algorithm.
- 5
- 5
- Activity-Based Person Identification Using Discriminative Sparse Projections and Orthogonal Ensemble Metric Learning5
An activity-based human identification approach using discriminative sparse projections (DSP) and orthogonal ensemble metric learning (OEML) to identify people from more general types of human activities such as eating, drinking, running, and so on.
- 5
- 5
- Outlier-Aware Post-Training Quantization for Image Super-Resolution4
This work proposes a dual-region quantization strategy that partitions activations into an outlier region and a dense region, applying uniform quantization to each region independently to better balance bit-width allocation.
- 4
- 4
- Efficient Meshy Neural Fields for Animatable Human Avatars4
EMA is presented, a method that Efficiently learns Meshy neural fields to reconstruct animatable human Avatars and jointly optimizes explicit triangular canonical mesh, spatial-varying material, and motion dynamics, via inverse rendering in an end-to-end fashion.
- 4
- 4
- Breast Mass Detection in Mammograms via Blending Adversarial Learning4
A generative model is trained to simulate the growth of mass on normal tissue by blending mass patches into healthy breast images by regularizing the imbalanced data with synthetically generated abnormal samples to make the training of deep learning based mass detector more stable and the resulting model more robust.
- 4
- 4
- 4
- 3
- Towards provable probabilistic safety for scalable embodied AI systems3
This Perspective argues for a paradigm shift to provable probabilistic safety that integrates provable guarantees with progressive achievement toward a probabilistic safety boundary on overall system performance, and outlines a roadmap for provable probabilistic safety, along with corresponding challenges and potential solutions.
- OPONeRF: One-Point-One NeRF for Robust Few-shot Rendering3
Experimental results show that the OPONeRF outperforms state-of-the-art few-shot neural rendering approaches on various evaluation metrics through benchmark experiments and cross-scene evaluations and incorporating the idea of One-Point-One NeRF into other advanced baseline methods.
- Vision Generalist Model: A Survey3
This paper provides a comprehensive overview of the vision generalist models, delving into their characteristics and capabilities within the field, and digs into the design of frameworks that have been proposed in existing research while also introducing the techniques employed to enhance their performance.
- 3
- 3D Finger Rotation Estimation from Fingerprint Images3
This paper proposes to estimate complete 3D relative finger angles based on two fingerprint images, which incorporate more information with a higher image resolution than capacitive images, and reveals that this method outperforms previous approaches with absolute finger angle models.
- Learning Accurate Performance Predictors for Ultrafast Automated Model Compression3
Experimental results on image classification and object detection show that the proposed SeerNet method achieves competitive accuracy-complexity trade-offs with significant reduction of the search cost.
- 3
- 3
- Left Atrial Appendage Neck Modeling for Closure Surgery3
The approach can precisely and non-invasively model the left atrial appendage neck for assisting closure surgeries and successfully solved problems including the auto-detection of the boundary points of left atrian appendage ostium, the establishment of the standard coordinate system and the polychromatic display of occlusive tension on the neck.
- Regularization techniques for high-dimensional data analysis3
This paper presents a meta-modelling architecture suitable for smart grids that was developed at the Tsinghua National Laboratory for Information Science and Technology (TNList) in China.
- 3
- 3
- Efficient Sparsity Estimation via Marginal-Lasso Coding3
The proposed method is more robust than the conventional marginal regression based methods and an adaptive regularization parameter selection scheme and a dictionary learning method incorporated with the proposed sparsity estimation algorithm.
- 3
- 3
- Human-Centric Indoor Environment Modeling from Depth Videos3
An approach to extract physical constraints from human poses and motion to better recover the spatial layout and model objects inside of indoor environments from depth videos is developed.
- 3
- 3
- 2
- 2
- InstaRevive: One-Step Image Enhancement via Dynamic Score Matching2
The proposed InstaRevive framework employs score-based diffusion distillation to harness potent generative capability and minimize the sampling steps, unveiling the compelling efficacy and efficiency of InstaRevive in delivering high-quality and visually appealing results.
- 2
- 2
- UniqueSplat: View-Conditioned 3D Gaussian Splatting for Generalizable 3D Reconstruction2
This paper proposes a two-branch view-conditioned hyperNetwork to simultaneously learn view-agnostic embeddings and view-specific knowledge, which not only explores the shareable knowledge from various views, but also adapts the model to specific views at test time.
- OPONeRF: One-Point-One NeRF for Robust Neural Rendering2
Experimental results show that the OPONeRF outperforms state-of-the-art NeRFs on various evaluation metrics through benchmark experiments and cross-scene evaluations, and the efficacy of the proposed method is shown via experimenting on other existing generalization-based benchmarks and incorporating the idea of One-Point-One NeRF into other advanced baseline methods.
- GlobalMamba: Global Image Serialization for Vision Mamba2
A global image serialization method to transform the image into a sequence of causal tokens, which contain global information of the 2D image, and builds a vision mamba model with a causal input format based on the proposed global image serialization, which can better exploit the causal relations among image sequences.
- 2
- Learning Generalizable Mixed-Precision Quantization via Attribution Imitation2
A robust generalizable mixed-precision quantization method to smooth the attribution for rank error alleviation by hierarchical attribution partitioning and proposes dynamic capacity-aware attribution imitation to adjust the concentration degree of the attribution according to sample hardness, so that sufficient model capacity is achieved with full utilization for each image.
- 2
- Category-level Shape Estimation for Densely Cluttered Objects2
This framework partitions each object in the clutter via the multi-view visual information fusion to achieve high segmentation accuracy, and the instance shape is recovered by deforming the category templates with diverse geometric transformations to obtain strengthened generalization ability.
- 2
- 2
- 2
- Metric Learning for Facial Kinship Verification2
This chapter introduces a neighborhood repulsed correlation metric learning (NRCML) method for facial kinship verification, and introduces a discriminative multi-metric learning (DMML), which makes use of multiple face descriptors, so that complementary and discrim inative information is exploited for verification.
- Video-Based Facial Kinship Verification2
This chapter presents a new video face dataset called Kinship Face Videos in the Wild (KFVW) which were captured in wild conditions for the video-based kinship verification study, as well as the standard benchmark, and employs the benchmark to evaluate and compare the performance of several state-of-the-art metric learning-based kinSHIP verification methods.
- Feature Learning for Facial Kinship Verification2
This chapter introduces a compact binary face descriptor (CBFD) method which learns face descriptors directly from raw pixels and presents a prototype-based discriminative feature learning (PDFL) method to learn mid-level discrim inative features with low-level descriptor for kinship verification.
- 2
- 2
- Fast object recognition using local scale-invariant features2
Methods for generating local scale invariant features are introduced for object recognition using hybrid multi-scale representation which has integrated the advantages of pyramids and scale-space representation.
- 1
- 1
- Learning Extensible Series-Parallel Lookup Tables for Efficient Image Super-Resolution1
This paper proposes SPLUT with learnable index quantizers, named LISPLUT, which automatically adjusts index distributions by non-uniformly allocating the quantization thresholds according to the input density distribution with minimal storage and computational overhead.
- 1
- 1
- 3D Small Object Detection with Dynamic Spatial Pruning1
A high-resolution multi-level detector with dynamic spatial pruning named DSPDet3D, which detects objects from large to small by iterative upsampling and meanwhile prunes the spatial representation of the scene at regions where there is no smaller object to be detected in higher resolution.
- 1
- 1
- Domain-Transferred Face Augmentation Network1
Experiments show that DotFAN is beneficial for augmenting small face datasets to improve their within-class diversity so that a better face recognition model can be learned from the augmented dataset.
- 1
- 1
- Detection of Substances in the Left Atrial Appendage by Spatiotemporal Motion Analysis Based on 4D-CT1
A new approach for the detection of substances in the left atrial appendage by spatiotemporal motion analysis and make a detailed judgment and analysis of spatial distribution and classification of most objects in the LAA is presented.
- 1
- 1
- 1
- Multiscale recurrent regression networks for face alignment1
This paper designs a recurrent regression network architecture, where the parameters across different stages are shared to memorize the shape residual descents between the initial shape and the ground-truth, and learns to exploit the context-aware information from multiscale face inputs in a coarse-to-fine manner.
- 1
- 1
- 1
- Gait Recognition Based on Independent Component Analysis and Information Fusion1
The experimental results show that the proposed gait recognition method has good recognition performance and is evaluated on the National Laboratory of Pattern Recognition and Xi'an University of Technology gait database and the correct recognition rate is relatively high.
- 1
- Test Time Adaptation with An Explicit Geometric Bridge for 3D Point Cloud Segmentation–
A category-balanced sampling mechanism that dynamically defines sampling thresholds based on class frequencies, which results in a similar proportion of classes in both easy and hard samples, thereby improving the issue of class imbalance.
- Scale-adaptive multi-predator network for camouflaged object detection–
This paper reformulates COD as a point-to-set matching problem and classify each pixel based on its distance with a set of dynamic query embeddings, and employs a confidence-based query selection method to enable each query to adaptively focus on different scales of the image for better identification of the camouflaged object.
- XMask3D++: Cross-Modal Mask Reasoning for Open Vocabulary 3D Segmentation–
This work developed a mask generator based on the denoising UNet from a pre-trained diffusion model, leveraging its capability for precise textual control over dense pixel representations and enhancing the open-world adaptability of the generated masks, and introduces a 2D-to-3D mask adaptation module to improve the quality of back-projected 3D masks.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Introduction to Facial Kinship Verification–
In this chapter, the background of facial kinship verification is introduced and the state-of-the-art of facial relatives verification is reviewed, to outline the organization of the book.
- –
- –
- –
- Simultaneous Audio Source Localization and Microphone Placement–
This paper presents a simple yet effective simultaneous audio source location and microphone placement approach to obtain the position information of stationary sound source by fusing geometric properties of the sliding microphone-pair system with an error theory analysis.
- –
- –
- –
- –
- –
- –
- –
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.