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
- Ying Nian WuSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works470 from public data
- 1,574
- 853
- 756
- 548
- 532
- 3D Hand Shape and Pose Estimation From a Single RGB Image517
This work proposes a Graph Convolutional Neural Network (Graph CNN) based method to reconstruct a full 3D mesh of hand surface that contains richer information of both 3D hand shape and pose and proposes a weakly-supervised approach by leveraging the depth map as a weak supervision in training.
- Robust hand gesture recognition based on finger-earth mover's distance with a commodity depth camera455
A novel distance metric for hand dissimilarity measure, called Finger-Earth Mover's Distance (FEMD), which only matches fingers while not the whole hand shape, can better distinguish hand gestures of slight differences.
- MixSTE: Seq2seq Mixed Spatio-Temporal Encoder for 3D Human Pose Estimation in Video444
This work proposes MixSTE (Mixed Spatio-Temporal Encoder), which has a temporal transformer block to separately model the temporal motion of each joint and a spatial transformer block for inter-joint spatial correlation, and extends from the central frame to entire frames of the input video, thereby improving the coherence between the input and output sequences.
- Track to Detect and Segment: An Online Multi-Object Tracker431
A new online joint detection and tracking model, TraDeS (TRAck to DEtect and Segment), exploiting tracking clues to assist detection end-to-end, infers object tracking offset by a cost volume, which is used to propagate previous object features for improving current object detection and segmentation.
- 357
- 354
- 347
- 333
- 327
- Robust hand gesture recognition with kinect sensor323
This work presents a hand gesture recognition system with Kinect sensor, which operates robustly in uncontrolled environments and is insensitive to hand variations and distortions, and applies a novel shape distance metric called Finger-Earth Mover's Distance (FEMD) forHand gesture recognition.
- Weakly-Supervised 3D Hand Pose Estimation from Monocular RGB Images320
A weakly-supervised method, adaptating from fully-annotated synthetic dataset toWeakly-labeled real-world dataset with the aid of a depth regularizer, which generates depth maps from predicted 3D pose and serves as weak supervision for3D pose regression.
- Abnormal event detection in crowded scenes using sparse representation320
This work proposes the sparse reconstruction cost (SRC) over the normal dictionary to measure the normalness of the testing sample, and introduces the prior weight of each basis during sparse reconstruction, which is more robust than other outlier detection criteria.
- 318
- 312
- 303
- Robust 3D Hand Pose Estimation in Single Depth Images: From Single-View CNN to Multi-View CNNs301
This work proposes to first project the query depth image onto three orthogonal planes and utilize these multi-view projections to regress for 2D heat-maps which estimate the joint positions on each plane to produce final 3D hand pose estimation with learned pose priors.
- PointCloud Saliency Maps284
A novel way of characterizing critical points and segments to build point-cloud saliency maps is proposed, and each saliency score can be efficiently measured by the corresponding gradient of the loss w.r.t the point under the spherical coordinates.
- 263
- Clustering Driven Deep Autoencoder for Video Anomaly Detection257
A novel convolution autoencoder architecture is designed to separately capture spatial and temporal informative representation to separately capture spatial and temporal informative representation for video anomaly detection.
- Multi-stream CNN: Learning representations based on human-related regions for action recognition247
A human-related multi-stream CNN (HR-MSCNN) architecture that encodes appearance, motion, and the captured tubes of the human- related regions is introduced that achieves state-of-the-art results on these four datasets.
- 239
- Video anomaly detection with spatio-temporal dissociation237
A novel convolution autoencoder architecture that can dissociate the spatio-temporal representation to separately capture the spatial and the temporal information is explored, since abnormal events are usually different from the normality in appearance and/or motion behavior.
- Bi-box Regression for Pedestrian Detection and Occlusion Estimation235
A novel approach to simultaneous pedestrian detection and occlusion estimation is proposed by regressing two bounding boxes to localize the full body as well as the visible part of a pedestrian respectively by learning a deep convolutional neural network consisting of two branches.
- 235
- 225
- 223
- Anomalous video event detection using spatiotemporal context223
Compared to other anomalous video event detection approaches that analyze object trajectories only, this work proposes a context-aware method to detect anomalies that is computationally efficient and can infer complex rules.
- Online Processing Algorithms for Influence Maximization201
This paper proposes a new algorithm for OPIM with both superior empirical effectiveness and strong theoretical guarantees, and demonstrates that its solutions outperform the state of the art for both OPIM and conventional influence maximization.
- 199
- Location Discriminative Vocabulary Coding for Mobile Landmark Search191
A Location Discriminative Vocabulary Coding (LDVC) scheme, which achieves extremely low bit rate query transmission, discriminative landmark description, as well as scalable descriptor delivery in a unified framework is presented.
- A2J: Anchor-to-Joint Regression Network for 3D Articulated Pose Estimation From a Single Depth Image190
The proposed 3D articulated pose estimation paradigm is different from the state-of-the-art encoder-decoder based FCN, 3D CNN and point-set based manners and 2D CNN is used as backbone network to drive A2J, without using time-consuming 3D convolutional or deconvolutional layers.
- 188
- Point-to-Point Regression PointNet for 3D Hand Pose Estimation185
The proposed Point-to-Point Regression PointNet directly takes the 3D point cloud as input and outputs point-wise estimations, i.e., heat-maps and unit vector fields on the point cloud, representing the closeness and direction from every point in the pointCloud to the hand joint.
- Large Margin Multi-metric Learning for Face and Kinship Verification in the Wild179
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.
- Human pose estimation and its application to action recognition: A survey176
This survey focuses on 2D skeleton-based action recognition where the human poses are estimated from regular RGB images instead of depth images, and summarizes the performance of recent action recognition methods that use pose estimated from color images as input, showing that there is much room for improvements in this direction.
- 170
- 168
- Discriminative Orderlet Mining for Real-Time Recognition of Human-Object Interaction160
This paper presents a novel visual representation, called orderlets, for real-time human action recognition with depth sensors, that is insensitive to small noise since an orderlet only depends on the comparative relationship among individual features.
- Two-Stream Consensus Network for Weakly-Supervised Temporal Action Localization158
A Two-Stream Consensus Network (TSCN) to simultaneously address weakly-supervised Temporal Action Localization challenges and a new attention normalization loss to encourage the predicted attention to act like a binary selection, and promote the precise localization of action instance boundaries.
- A survey of variational and CNN-based optical flow techniques154
This paper provides a systematic review of recent optical flow techniques with a focus on the variational method and approaches based on Convolutional Neural Networks (CNNs) and discusses their potential and current limitations.
- Joint-Bone Fusion Graph Convolutional Network for Semi-Supervised Skeleton Action Recognition152
A novel correlation-driven joint-bone fusion graph convolutional network (CD-JBF-GCN) as an encoder and a pose prediction head as a decoder to achieve semi-supervised learning and allows the network to learn motion representation from the unlabeled data, which is essential for action recognition.
- 150
- 149
- 143
- 141
- 137
- 135
- Model-based 3D Hand Reconstruction via Self-Supervised Learning134
This work proposes S2HAND, a self-supervised 3D hand reconstruction network that can jointly estimate pose, shape, texture, and the camera viewpoint and utilizes the consistency between 2D and 3D representations and a set of novel losses to rationalize outputs of the neural network.
- Hand-Transformer: Non-Autoregressive Structured Modeling for 3D Hand Pose Estimation128
This work proposes to leverage the Transformer architecture with a novel non-autoregressive structured decoding mechanism, and demonstrates the model’s attractiveness over multiple challenging hand pose datasets, comparing with several state-of-the-art methods.
- 128
- 3DV: 3D Dynamic Voxel for Action Recognition in Depth Video127
3D dynamic voxel (3DV) is proposed as a novel 3D motion representation manner that is lightweight and effective for deep feature learning towards point set and split the depth video into temporal segments and encoded in 3DV integrally.
- 119
- Kernel Cross-Correlator116
By introducing the kernel trick, the KCC extends the linear cross-correlation to non-linear space, which is more robust to signal noises and distortions, and eliminates direct calculation of kernel vectors, thus achieves better performance yet still with a reasonable computational cost.
- SLCGC: A lightweight Self-supervised Low-Pass Contrastive Graph Clustering Network for Hyperspectral Images110
This approach begins with homogeneous region generation, which aggregates pixels into spectrally consistent regions to preserve local spatial-spectral coherence while drastically reducing graph complexity, and introduces a low-pass graph denoising mechanism to suppress high-frequency noise in the graph topology.
- 108
- SPAGAN: Shortest Path Graph Attention Network107
This paper proposes a novel GCN model, which it is term as Shortest Path Graph Attention Network (SPAGAN), which conducts path-based attention that explicitly accounts for the influence of a sequence of nodes yielding the minimum cost, or shortest path, between the center node and its higher-order neighbors.
- 104
- 104
- 103
- Real-Time 3D Hand Pose Estimation with 3D Convolutional Neural Networks101
A novel 3D CNN-based method that can capture the 3D spatial structure of the hand and accurately regress full 3D hand pose in a single pass is presented and is fast as the implementation runs at over 91 frames per second on a standard computer with a single GPU.
- Motion-Driven Visual Tempo Learning for Video-Based Action Recognition98
This work proposes a Temporal Correlation Module (TCM), which can be easily embedded into the current action recognition backbones in a plug-in-and-play manner, to extract action visual tempo from low-level backbone features at single-layer remarkably.
- 96
- 96
- 95
- 95
- 94
- 94
- Fast and robust short video clip search using an index structure93
An index structure-based method to fast and robustly search short video clips in large video collections and develops a set of spatial-temporal features in order to globally capture the pattern of a short video clip and combine it with the color range feature to form video signatures.
- Cross-Modal Video Moment Retrieval with Spatial and Language-Temporal Attention92
The proposed two attention sub-networks can recognize the most relevant objects and interactions in the video, and simultaneously highlight the keywords in the query, and demonstrate the effectiveness of the model as compared to state-of-the-art methods.
- 3D fingertip and palm tracking in depth image sequences90
Quantitative results on synthetic depth sequences show the proposed scheme can track the fingertips quite accurately, and its capabilities are further demonstrated through a real-life human-computer interaction application.
- NeuRBF: A Neural Fields Representation with Adaptive Radial Basis Functions89
A novel type of neural fields that uses general radial bases for signal representation that achieves state-of-the-art rendering quality, with small model size and comparable training speed by marrying adaptive radial bases with grid-based ones.
- 87
- 86
- ACSNet: Action-Context Separation Network for Weakly Supervised Temporal Action Localization85
An Action-Context Separation Network (ACSNet) that explicitly takes into account context for accurate action localization is introduced and extended labels with auxiliary context categories are introduced to facilitate the learning of action-context separation.
- 85
- Joint Hand-Object 3D Reconstruction From a Single Image With Cross-Branch Feature Fusion84
This work proposes to consider hand and object jointly in feature space and explore the reciprocity of the two branches in cross-branch feature fusion architectures with MLP or LSTM units, and significantly outperforms existing approaches in terms of the reconstruction accuracy of objects.
- Randomized Spatial Partition for Scene Recognition84
The objective is to characterize the image layout by various patterns, in the form of randomized spatial partition (RSP), which makes it possible to mine the most descriptive image layout pattern for each category of scenes, and then combine them by training a discriminative classifier, i.e., the proposed ORSP classifier.
- Learning Diverse Stochastic Human-Action Generators by Learning Smooth Latent Transitions83
This paper focuses on skeleton-based action generation and proposes to model smooth and diverse transitions on a latent space of action sequences with much lower dimensionality, and is learned with a bi-directional generative-adversarial-net framework.
- 83
- 82
- Forest R-CNN: Large-Vocabulary Long-Tailed Object Detection and Instance Segmentation80
Prior knowledge of the relations among object categories is exploited to cluster fine-grained classes into coarser parent classes, and construct a classification tree that is responsible for parsing an object instance into a fine- grained category via its parent class.
- Robust 3D Hand Pose Estimation From Single Depth Images Using Multi-View CNNs80
A novel multi-view convolutional neural network (CNN)-based approach for 3D hand pose estimation that is superior to the state-of-the-art methods on two challenging data sets and validates that the proposed approach has good generalization ability.
- Product Quantization Network for Fast Image Retrieval80
Through the proposed product quantization network, the author can obtain a discriminative and compact image representation in an end-to-end manner, which further enables a fast and accurate image retrieval.
- 80
- 78
- Consistent 3D Hand Reconstruction in Video via Self-Supervised Learning77
A self-supervised 3D hand reconstruction model that can jointly estimate pose, shape, texture, and the camera viewpoint from a single RGB input through the supervision of easily accessible 2D detected keypoints is presented.
- Deformable Pose Traversal Convolution for 3D Action and Gesture Recognition77
A Deformable Pose Traversal Convolution Network is proposed that applies one-dimensional convolution to traverse the 3D pose for its representation and optimizes the convolution kernel for each joint, by considering contextual joints with various weights.
- 77
- 76
- 75
- Learning LBP structure by maximizing the conditional mutual information74
An incremental Maximal-Conditional-Mutual-Information scheme for LBP structure learning to handle pixel correlation has demonstrated a superior performance over the state-of-the-arts results on classifying both spatial patternssuch as texture classification, scene recognition and face recognition, and spatial-temporal patterns such as dynamic texture recognition.
- 73
- Human–Robot Interaction by Understanding Upper Body Gestures72
The integration of a handshake within the HRI system illustrates the effectiveness of the proposed online generation method, and the experimental results demonstrate that the approach can recognize the upper body gestures with high accuracy in real time.
- 71
- HOT-Net: Non-Autoregressive Transformer for 3D Hand-Object Pose Estimation70
The proposed Hand-Object Transformer Network (HOT-Net) removes the Transformer's dependence on previously generated results and explicitly feeds a reference 3D hand-object pose into the decoding process to provide equivalent target pose patterns for parallely localizing each 3D keypoint.
- 70
- Self-Mimic Learning for Small-scale Pedestrian Detection67
An in-depth analysis of the small-scale pedestrian detection problem is conducted, which reveals that weak representations of small- scale pedestrians are the main cause for a classifier to miss them and proposes a novel Self-Mimic Learning (SML) method to improve the detection performance.
- 66
- 66
- 63
- Propagative Hough Voting for Human Activity Recognition63
The proposed activity recognition method does not rely on human detection and tracking, and can well handle the scale and intra-class variations of the activity patterns, and outperforms the state-of-the-art techniques when there is sufficient training data.
- 62
- 61
- CATS: Co-saliency Activated Tracklet Selection for Video Co-localization60
This work proposes to leverage co-saliency activated tracklets to address the challenge of jointly localizing common objects across videos and shows that the proposed method outperforms state-of-the-art methods.
- 60
- Predicting human activities using spatio-temporal structure of interest points60
A new random forest structure is proposed, called multi-class balanced random forest, which makes a good trade-off between the balance of the trees and the discriminative abilities and significantly outperforms the state of the arts for the human activity prediction problem.
- 59
- 59
- Fusion of 3D-LIDAR and camera data for scene parsing59
An approach which performs scene parsing and data fusion for a 3D-LIDAR scanner (Velodyne HDL-64E) and a video camera and the fused results are more reliable than that acquired via analysis of only images or Velodyne data.
- Model-based hand pose estimation via spatial-temporal hand parsing and 3D fingertip localization59
Quantitative experiments on synthetic data show the proposed hand pose estimation scheme can accurately capture the natural hand motion, and a simulated water-oscillator application is built to demonstrate the effectiveness of the proposed method in human-computer interaction scenarios.
- An efficient and effective hop-based approach for influence maximization in social networks58
This paper proposes hop-based algorithms that can be easily applied to billion-scale networks under the commonly used Independent Cascade and Linear Threshold influence diffusion models and provides provable data-dependent approximation guarantees for these approaches.
- Measuring Generalisation to Unseen Viewpoints, Articulations, Shapes and Objects for 3D Hand Pose Estimation Under Hand-Object Interaction57
A public challenge to evaluate the abilities of current 3D hand pose estimators (HPEs) to interpolate and extrapolate the poses of a training set, with dramatically improved accuracy over the baseline.
- Invariant multi-scale descriptor for shape representation, matching and retrieval57
A novel invariant multi-scale descriptor to represent shape contour is proposed that is invariant to rotation, scale variation, intra-class variation, articulated deformation and partial occlusion, and is robust to noise as well.
- Bayesian Uncertainty Matching for Unsupervised Domain Adaptation56
This work uses a Bayesian neural network to quantify prediction uncertainty of a classifier and proposes an approximate joint distribution matching scheme by exploiting prediction uncertainty, which effectively alleviates label distribution mismatching in source and target data.
- 55
- 55
- 54
- A complete and fully automated face verification system on mobile devices54
This paper presents a natural and non-intrusive way to secure mobile devices, i.e. a complete and fully automated face verification system that consists of three sub-systems: face detection, alignment and verification, which locates the eyes at a much higher precision than Adaboost face/eye detector.
- From frequent itemsets to semantically meaningful visual patterns54
The novelty of this work lies in the following new contributions: a principled solution to the discovery of meaningful itemsets based on frequent itemset mining; a self-supervised clustering scheme of the high-dimensional visual features by feeding back discovered patterns to tune the similarity measure through metric learning; and a pattern summarization method that deals with the measurement noises brought by the image data.
- 53
- 52
- AR in Hand52
This demo shows the possibility to interact with 3D contents with bare hands on wearable devices by two Augmented Reality applications, including virtual teapot manipulation and fountain animation in hand.
- Influence Maximization Meets Efficiency and Effectiveness51
This paper proposes hopbased algorithms that can easily scale to millions of nodes and billions of edges and can provide certain theoretical guarantees in relation to influence Maximization.
- Representative Selection with Structured Sparsity51
Experiments on image and video data validate the effectiveness of the method in finding exemplars with diversity and representativeness and demonstrate its robustness to outliers.
- Efficient Video Instance Segmentation via Tracklet Query and Proposal50
This paper proposes EfficientVIS, a fully end-to-end framework with efficient training and inference that associate and segment regions-of-interest (RoIs) across space and time by an iterative query-video interaction and proposes a correspondence learning that makes tracklets linking between clips end- to-end learnable.
- Federated Learning With Privacy-Preserving Ensemble Attention Distillation50
This work proposes a privacy-preserving FL framework leveraging unlabeled public data for one-way offline knowledge distillation in this work and achieves very competitive performance with more robust privacy preservation based on extensive experiments on image classification, segmentation, and reconstruction tasks.
- Structure-Aware Human-Action Generation50
This work proposes a variant of GCNs to leverage the powerful self-attention mechanism to adaptively sparsify a complete action graph in the temporal space, and demonstrates the superiority of this method on two standard human action datasets compared with existing methods.
- 50
- Mining Motifs from Human Motion50
This paper proposes an efficient motif discovery method which can handle both spatia l and temporal variations of motion data, and uses locality-sensitive hashing (LSH) to find the approximate matches in a matching trellis.
- 49
- Exploiting Local Feature Patterns for Unsupervised Domain Adaptation48
Comparisons to the state-of-the-art unsupervised domain adaptation methods on two popular benchmark datasets demonstrate the superiority of this approach and its effectiveness on alleviating negative transfer.
- Survey on depth and RGB image-based 3D hand shape and pose estimation47
This work states that the existence of complicated hand articulation, depth and scale ambiguities, occlusions, and the availability of large-scale annotated hand datasets and the rapid developments of deep neural networks (DNNs) means that accurate and rapid hand shape and pose estimation is needed.
- Fusing disparate object signatures for salient object detection in video47
A novel spatiotemporal saliency model for object detection in videos that aims to use object signatures which can be identified by any kinds of object segmentation methods and outperforms existing state-of-the-art approaches is presented.
- 47
- Entropic image thresholding based on GLGM histogram47
GLGM (gray-level & gradient-magnitude) histogram is proposed as a novel image histogram for thresholding and employs the Fibonacci quantized gradient magnitude to characterize spatial information effectively.
- 46
- 46
- 46
- Uncertainty-aware State Space Transformer for Egocentric 3D Hand Trajectory Forecasting45
An uncertainty-aware state space Transformer (USST) is proposed that takes the merits of the attention mechanism and aleatoric uncertainty within the framework of the classical state-space model and can be further enhanced by the velocity constraint and visual prompt tuning on large vision transformers.
- 45
- Towards Generic Image Manipulation Detection with Weakly-Supervised Self-Consistency Learning44
Weakly-supervised image manipulation detection, such that only binary image-level labels (authentic or tampered with) are required for training purpose, and two consistency properties are learned: multi-source consistency (MSC) and inter-patch consistency (IPC).
- 44
- NeuLF: Efficient Novel View Synthesis with Neural 4D Light Field44
This method can render novel views by sampling rays and querying the color for each ray from the network directly, thus enabling high-quality light field rendering with a sparser set of training images and enabling applications such as auto refocus.
- 44
- 42
- 42
- 41
- 41
- 40
- Towards Real-Time Eyeblink Detection in the Wild: Dataset, Theory and Practices40
A modified LSTM model able to capture the multi-scale temporal information is proposed, and a feature extraction approach that reveals the appearance and motion characteristics simultaneously is also proposed to verify eyeblink.
- Towards Profit Maximization for Online Social Network Providers40
A seed set to optimize a new profit metric that combines the benefit of influence spread with the cost of influence propagation for the OSN provider is found and can achieve high approximation guarantees and significantly outperform the baseline algorithms, including state-of-the-art influence maximization algorithms.
- 39
- Barehanded music39
Preliminary results show that the efficient data-driven approach to track fingertip and detect finger tapping for virtual piano using an RGB-D camera can recognize most of the beginner's piano-playing gestures in realtime for soothing rhythms.
- Learning a robust representation via a deep network on symmetric positive definite manifolds38
This paper proposes to aggregate deep convolutional features into an SPD matrix representation through the SPD generation and the SPD transformation under an end-to-end deep network, and shows that the method notably outperforms state-of-the-art methods on visual classification.
- 38
- 38
- 38
- 37
- Minimum Near-Convex Shape Decomposition37
The minimum near-convex decomposition (MNCD) to decompose arbitrary shapes into minimum number of "near-convex" parts is proposed and outperforms the state-of-the-art results without introducing redundant parts and thus leads to robust shape representation.
- Middle-Level Representation for Human Activities Recognition: The Role of Spatio-Temporal Relationships37
It is demonstrated experimentally that the middle-level representation combined with a χ2-SVM classifier equals to or outperforms the state-of-the-art results on these dataset.
- 37
- Temporal Distinct Representation Learning for Action Recognition36
This paper designs a sequential channel filtering mechanism, i.e., Progressive Enhancement Module (PEM), to excite the discriminative channels of features from different frames step by step to avoid repeated information extraction and achieves visible improvements over the best competitor by 2.4% and 1.3%, respectively.
- 36
- 35
- 35
- Learning to Integrate Occlusion-Specific Detectors for Heavily Occluded Pedestrian Detection34
This work categorizes occlusions based on how pedestrian examples are occluded into K groups and adopts an L1-norm linear support vector machine (SVM) to select and fuse occlusion-specific detectors for the K classifiers simultaneously.
- 33
- Fast and Robust Short Video Clip Search for Copy Detection33
This paper designs a novel global visual feature (a fixed-size 144-d signature) combining the spatial-temporal and the color range information and achieves an efficient and robust solution for video copy detection.
- Human-virtual human interaction by upper body gesture understanding32
An effective and real-time human gesture recognition algorithm is proposed based on the posture data from the CyberGlove II and the Kinect, and it is demonstrated that the algorithm can recognize human gestures with high accuracy in real time.
- Weakly Supervised Temporal Action Localization Through Learning Explicit Subspaces for Action and Context31
This work introduces a framework that learns two feature subspaces respectively for actions and their context that outperforms state-of-the-art WS-TAL methods on three benchmarks, and leverages the predictions from both spatial and temporal streams for snippets grouping.
- Multi-label learning of part detectors for occluded pedestrian detection31
A multi-label learning approach to jointly learn part detectors to capture partial occlusion patterns to reduce the computational cost of applying these part detectors for pedestrian detection and show better performance when used as a pedestrian detector individually.
- Hough Forest With Optimized Leaves for Global Hand Pose Estimation With Arbitrary Postures31
A novel algorithm to optimize the leaf weights in a Hough forest to assist global hand pose estimation with a single depth camera and demonstrates improved robustness compared to several other recent hand tracking systems from both industry and academy.
- Query Adaptive Instance Search using Object Sketches31
The proposed query adaptive sketch-based object search is formulated as a sub-graph selection problem, which can be solved by maximum flow algorithm and can accurately locate the small target objects in cluttered background or densely drawn deformation intensive cartoon images.
- Mining GPS traces and visual words for event classification31
Experimental results using leave-one-out cross validation have shown the synergy of all of the components in the proposed approach to event classification, which performs multiclass AdaBoost to solve the multiclass problem.
- 31
- 30
- 30
- Optical flow for video super-resolution: a survey30
An in-depth study of the deep learning based video super-resolution method, where some representative algorithms are analyzed and compared and some promising research directions and open issues are highlighted.
- Interactive visual object search through mutual information maximization30
This work forms the object search as the problem of finding subimages with the largest mutual information toward the query object, and proposes a discriminative matching using both positive and negative queries to obtain the mutual information score.
- 29
- Non-iterative SLAM29
This method is the first non-iterative and online trainable approach for data association in visual SLAM that runs at a faster speed and obtains 3-D maps with higher resolution yet still with comparable accuracy.
- 29
- Hierarchical domain adaptation with local feature patterns28
The proposed method achieves more reliable feature transfer by consistently outperforming state-of-the-art models and the learned domain-invariant features generalize well to novel domains.
- Efficient Object Instance Search Using Fuzzy Objects Matching28
A Fuzzy Objects Matching (FOM) framework to effectively and efficiently capture the relevance between the query object and reference images in the dataset is proposed and significantly outperforms the state-of-the-art methods in precision with less memory and computational cost on three public datasets.
- 28
- Visual pattern discovery in image and video data: a brief survey27
A review of the major progress in visual pattern discovery is provided, categorizing the existing methods into two groups: bottom‐up pattern discovery and top‐down pattern modeling.
- High Fidelity 3D Hand Shape Reconstruction via Scalable Graph Frequency Decomposition26
A frequency split network to generate 3D hand mesh using different frequency bands in a coarse-to-fine manner to capture high-frequency personalized details, and proposes a novel frequency decomposition loss to supervise each frequency component.
- Product Quantization Network for Fast Visual Search26
Comprehensive experiments conducted on multiple public benchmark datasets demonstrate the state-of-the-art performance of the proposed PQN, RPQN and TPQN in fast image and video retrieval.
- 26
- 26
- Mining and cropping common objects from images26
This work describes each image as a collection of visual primitives and proposes a novel bottom-up approach to gradually prune local primitives to recover the whole common object, designed to accelerate the image data mining process.
- 25
- 25
- 25
- Chain-of-Look Prompting for Verb-centric Surgical Triplet Recognition in Endoscopic Videos24
This work casts the problem of surgical triplet recognition as visual prompt generation from large-scale vision-language (VL) models, and explicitly decomposes the task into a series of video reasoning processes, which captures the visual reasoning processes underlying surgical activities and achieves better performance compared to the state-of-the-art methods.
- Modelling Multi-Party Interactions among Virtual Characters, Robots, and Humans24
The challenges in autonomous multi-party interaction among virtual characters, human-like robots, and real participants are discussed, and a prototype system to study these challenges is described.
- 24
- 23
- 23
- 22
- Efficient-NeRF2NeRF: Streamlining Text-Driven 3D Editing with Multiview Correspondence-Enhanced Diffusion Models22
This work proposes that by incorporating correspondence regularization into diffusion models, the process of 3D editing can be significantly accelerated by leveraging this multiview consistency and can edit 3D content at a much faster speed.
- 22
- Positive and Unlabeled Learning for Anomaly Detection with Multi-features22
This work introduces a new framework based on Positive and Unlabeled (PU) Learning using multi- features to detect anomalies and extends previous PU learning methods to better address unbalanced class problem which is typical for anomaly detection.
- Boosting cross-media retrieval via visual-auditory feature analysis and relevance feedback22
This paper builds multimodal representation via statistical canonical correlation between image and audio feature matrices, and defines cross-media distance metric for similarity measure, and proposes optimization strategy based on relevance feedback, which fuses short-term learning results and long-term accumulated knowledge into the objective function.
- Anomaly Detection and Generation with Diffusion Models: A Survey21
This survey comprehensively review anomaly detection and generation with diffusion models (ADGDM), presenting a tutorial-style analysis of the theoretical foundations and practical implementations and spanning images, videos, time series, tabular, and multimodal data.
- Scene reconstruction techniques for autonomous driving: a review of 3D Gaussian splatting21
This paper comprehensively reviews and summarizes the existing related research on 3D Gaussian Splatting and the research frontier directions of 3D GS in autonomous driving, including structure optimization, 4D scene reconstruction, and cross-domain research.
- Handling Difficult Labels for Multi-label Image Classification via Uncertainty Distillation21
This work proposes to calibrate the model, which not only predicts the labels but also estimates the uncertainty of the prediction, and locates the image regions that cause such uncertain predictions through uncertainty attention distillation.
- 21
- 21
- 21
- Real-time human action search using random forest based hough voting21
This paper proposes a novel framework for action retrieval which performs pattern matching at subvolume level and is very efficient in handling large corpus of videos and presents action search experiments on challenging datasets to validate the efficiency and effectiveness of the system.
- 21
- KPB-SIFT21
The produced KPB-SIFT descriptor is more compact as compared to the state-of-the-art, does not require pre-training step needed by PCA based descriptors, and shows superior advantages in terms of distinctiveness, invariance to scale, and tolerance of geometric distortions.
- 20
- 20
- 20
- 19
- 19
- SOAR: Scene-debiasing Open-set Action Recognition19
A novel method, called Scene-debiasing Open-set Action Recognition (SOAR), which features an adversarial scene reconstruction module and an adaptive adversarialscene classification module that prevents the decoder from reconstructing the video background given video features, and thus helps reduce the background information in feature learning.
- 19
- 19
- Parsing 3D motion trajectory for gesture recognition19
The proposed view invariant hierarchical parsing method for free form 3D motion trajectory representation is view-invariant in 3D space and is robust to variations of scale, temporary speed and partial occlusion.
- Tracking and fusion for multiparty interaction with a virtual character and a social robot19
This paper provides the details of the tracking and fusion component including speaker identification, addressee detection and a dynamic user entrance/leave mechanism based on user re-identification using a Kinect sensor.
- 19
- 19
- Speeding up spatio-temporal sliding-window search for efficient event detection in crowded videos19
This work proposes a heuristic branch-and-bound solution for event detection in videos by directly finding subvolumes whose scores are higher than the threshold, which dramatically reduces computational cost without significantly degrading the quality of the detection results.
- IDOL: Unified Dual-Modal Latent Diffusion for Human-Centric Joint Video-Depth Generation18
This work proposes a unified dual-modal U-Net, a parameter-sharing framework for joint video and depth denoising, wherein a modality label guides the denoising target, and cross-modal attention enables the mutual information flow.
- PREF: Predictability Regularized Neural Motion Fields18
The proposed framework PREF (Predictability REgularized Fields) achieves on par or better results than state-of-the-art neural motion field-based dynamic scene representation methods, while requiring no prior knowledge of the scene.
- 18
- 18
- 18
- 18
- 18
- dFLMoE: Decentralized Federated Learning via Mixture of Experts for Medical Data Analysis17
The proposed dFLMoE framework outperforms state-of-the-art approaches under both model homogeneity and heterogeneity settings and removes the dependency on the central server to enhance the robustness of the framework.
- 17
- Detecting spatiotemporal irregularities in videos via a 3D convolutional autoencoder17
A 3D fully convolutional autoencoder (3D-FCAE) that is trainable in an end-to-end manner to detect both temporal and spatiotemporal irregularities in videos using limited training data is introduced and shows good generalizability across multiple datasets.
- SLTFNet: A spatial and language-temporal tensor fusion network for video moment retrieval17
Experimental results on three public benchmark datasets show that the SLTF model significantly outperforms current state-of-the-art approaches, and the benefits produced by new technologies incorporated into SLTF are demonstrated.
- Quality of experience measurement for light field 3D displays on multilayer LCDs17
This work proposes three customized virtual models, which are the USAF‐E model, the view angle model, and the concave/convex object model for accurate measurement of spatial resolution, viewing angle, and depth resolution for MLLFDs.
- 17
- 17
- Saliency Density Maximization for Object Detection and Localization17
A fast approach to detect salient objects from images with extensive results on different types of saliency maps with a public dataset of five thousand images shows the advantages of the approach as compared to some state-of-the-art methods.
- 16
- FSC: Few-Point Shape Completion16
Few-point Shape Completion model is proposed, which contains a novel dual-branch feature extractor for handling extremely sparse inputs, coupled with an extensive branch for maximal point utilization with a saliency branch for dynamic importance assignment, and shows good gener-alizability to different object categories.
- Image Co-Skeletonization via Co-Segmentation16
This article explores a new joint processing topic: image co-skeletonization, which is defined as joint skeleton extraction of the foreground objects in an image collection, and proposes a coupled framework for co-Skeletonization and co-segmentation tasks to facilitate shape information discovery for this process through the co- Segmentation process.
- 16
- 16
- 16
- 16
- Salient object detection in videos by optimal spatio-temporal path discovery15
This work proposes to detect and track salient object simultaneously by finding a spatio-temporal path of the highest saliency density in the video by leveraging the motion coherence of videos.
- 15
- 15
- Texture or Semantics? Vision-Language Models Get Lost in Font Recognition14
Through extensive evaluation of various VLMs on font recognition tasks, the Font Recognition Benchmark (FRB) is introduced, a compact and well-structured dataset comprising 15 commonly used fonts that introduces a stroop effect that challenges model perception.
- 14
- 14
- 14
- 14
- Mining repetitive clips through finding continuous paths14
A novel method which translates repetitive clip mining to the continuous path finding problem in a matching trellis, where sequence matching can be accelerated by taking advantage of the temporal redundancies in the videos.
- 14
- Robust Knowledge Transfer via Hybrid Forward on the Teacher-Student Model13
A hybrid forward scheme is proposed in training the teacher-student models, alternately updating layer weights of the student model to ease the efforts of transferring knowledge from the teacher to the student network, robust to the gaps between their network architectures, domain data, and task definitions.
- 13
- 13
- 13
- 12
- 12
- Tensorized Projection for High-Dimensional Binary Embedding12
Tensorized Projection (TP) is proposed to decompose the projection matrix using Tensor-Train (TT) format, which is a chain-like representation that allows to operate tensor in an efficient manner, which can drastically reduce the computational complexity and memory cost.
- 12
- 12
- 12
- 12
- Neural Voting Field for Camera-Space 3D Hand Pose Estimation11
The proposed Neural Voting Field (NVF) fully models 3D dense local evidence and hand global geometry, helping to alleviate common 2D-to-3D ambiguities and outperforms existing state-of-the-art algorithms on FreiHAND dataset for camera-space 3D hand pose estimation.
- 11
- 11
- 11
- 11
- 11
- 11
- Finding spatio-temporal salient paths for video objects discovery11
Inspired by the observation that salient video objects usually appear in consecutive frames, the motion coherence of videos is leveraged into the path discovery and make the salient object detection more robust.
- Face and Facial Expressions Recognition and Analysis11
An extended noise-resistant LBP (ENRLBP) to capture line patterns and NRLBP and ENRLBP are validated extensively on face recognition, facial expression analysis and other recognition tasks and are shown more resistant to image noise compared with LBP, LTP and many other variants.
- 11
- Trends and Topics in Computer Vision11
This work learns 20 visual attributes and uses them in a zero-shot transfer learning experiment as well as to make visual connections between semantically unrelated object categories.
- 11
- Spectrum AUC Difference (SAUCD): Human-Aligned 3D Shape Evaluation10
An analytic metric named Spectrum Area Under the Curve Difference (SAUCD) is proposed that demonstrates better consistency with human evaluation, and outperforms previous 3D mesh metrics.
- Multimodal Co-training for Fake News Identification Using Attention-aware Fusion10
Anective attention-aware joint representation learning framework that learns the comprehensive data pattern by correlating each word in the text component to each potential object region in the image component by designing a novel multimodal co-training mechanism.
- 10
- 10
- 10
- Distributed Composite Quantization10
Experiments on ANN search and image retrieval tasks validate that the proposed DCQ significantly improves Composite Quantization in both efficiency and scale, while still maintaining competitive accuracy.
- One-stop measurement model for fast and accurate tensor display characterization10
An improved model, based on an earlier design concept, has been developed that not only can be used to measure spatial and angular cutoff frequencies, but also can enable comprehensive characterization of the display.
- Random Forest with Suppressed Leaves for Hough Voting10
The experimental results demonstrate that by suppressing unreliable leaf nodes, it not only improves prediction accuracy, but also reduces both prediction time cost and model complexity of the random forest.
- 10
- Non-rectangular Part Discovery for Object Detection10
A novel datadriven approach to discover non-rectangular parts by exploiting object structures is proposed to address the limitations of most DPM-based methods.
- Location Constrained Pixel Classifiers for Image Parsing with Regular Spatial Layout10
The proposed method utilizes the spatial layout of the image by building local pixel classifiers that are location constrained, i.e., trained with pixels from a local neighborhood region only, and works surprisingly well in different challenging image parsing problems, such as pedestrian parsing and object segmentation.
- Hand Parsing and Gesture Recognition with a Commodity Depth Camera10
This chapter presents a robust hand parsing scheme to obtain a high-level and discriminative representation of the hand from raw depth image, and develops a hand gesture recognition algorithm that achieves good performance compared to state-of-the-art methods.
- 9
- 9
- An Image-Based Approach to Detecting Structural Similarity Among Mixed Integer Programs9
In this paper, CCM structure is visualized using digital images, and computer vision techniques are used to detect latent structural features therein and measure similarity between images and, consequently, MIPs.
- 9
- 9
- 9
- 9
- Hand pose estimation by combining fingertip tracking and articulated ICP9
Experiments on both synthetic data and real-world sequences show the model-based framework for hand pose estimation scheme can accurately capture the natural hand motion.
- Harnessing Low-Frequency Neural Fields for Few-Shot View Synthesis8
This work harnesses low-frequency neural fields to regularize high- frequencies neural fields from overfitting to better address the problem of few-shot view synthesis and proposes a simple-yet-effective strategy for tuning the frequency to avoid overfitting few- shot inputs.
- Revisiting Modified Greedy Algorithm for Monotone Submodular Maximization with a Knapsack Constraint8
- 8
- First-Person Palm Pose Tracking and Gesture Recognition in Augmented Reality8
An Augmented Reality solution to allow users to manipulate and inspect 3D virtual objects freely with their bare hands on wearable devices is presented and a unified framework to jointly recover the 6D palm pose and recognize the hand gesture from the depth images is proposed.
- 8
- 8
- Activity recognition in unconstrained RGB-D video using 3D trajectories8
A novel RGB--D activity recognition approach that leverages the dense trajectory feature in RGB videos to recover the 3D trajectory of the tracked interest points, which captures important motion information along the depth direction.
- 8
- On preserving robustness-false alarm tradeoff in media hashing8
This paper specifically demonstrates the kind of robustness false alarm tradeoff achieved by existing methods and proposes a method for hashing that clearly improves this tradeoff.
- Benchmarking large and small MLLMs7
It is revealed that small MLLMs can achieve comparable performance to large models in specific scenarios but lag significantly in complex tasks requiring deeper reasoning or nuanced understanding, highlighting domains where even state-of-the-art models struggle.
- AMuSE: Adaptive Multimodal Analysis for Speaker Emotion Recognition in Group Conversations7
A Multimodal Attention Network (MAN) that captures cross-modal interactions at various levels of spatial abstraction by jointly learning its interactive bunch of mode-specific Peripheral and Central networks and helps the users in understanding the reasoning behind each emotion prediction made by the model via its Multimodal Explainability Visualization module.
- 7
- HYPERSPECTRAL IMAGE CLASSIFICATION WITH LOCALIZED SPECTRAL FILTERING-BASED GRAPH ATTENTION NETWORK7
The presented approach offers promising classification performance, that is, the GAT using t-SNE acquires superior performance than that of using PCA (principal component analysis), and also proves the great importance of combining spatial- and spectral information for hyperspectral image classification.
- 7
- Understanding Human-Object Interaction in RGB-D videos for Human Robot Interaction7
This work proposes to leverage human body pose as the context to achieve robust hand-held object detection in RGB-D videos and demonstrates a person can interact with a humanoid social robot with hand- held object such as a cell phone or a cup.
- Sound-event classification using pseudo-color CENTRIST feature and classifier selection7
This work proposes to extract pseudo-color CENTRIST features from the logarithm of Gammatone-like spectrogram to well classify the sound event under the unknown noise condition with a classifier-selection scheme, which automatically selects the most suitable classifier.
- 7
- 7
- UST-SSM: Unified Spatio-Temporal State Space Models for Point Cloud Video Modeling6
The Unified Spatio-Temporal State Space Model (UST-SSM) is proposed, which extends the latest advancements in SSMs to point cloud videos and introduces Spatial-Temporal Selection Scanning (STSS), which reorganizes unordered points into semantic-aware sequences through prompt-guided clustering, thereby enabling the effective utilization of points that are spatially and temporally distant yet similar within the sequence.
- 6
- 6
- Uncovering Malicious Accounts in Online Social Networks Using XGBoost and Graph Convolution Networks6
- 6
- 6
- 6
- 6
- 6
- 6
- Is My Object in This Video? Reconstruction-based Object Search in Videos6
This paper addresses the problem of video-level object instance search, which aims to retrieve the videos in the database that contain a given query object instance, and proposes the Reconstruction-based Object SEarch (ROSE) method, which characterizes a huge corpus of features of possible spatial-temporal locations in the video into the parameters of the reconstruction model.
- 6
- 6
- 6
- 6
- Arbitrary-Shape Object Localization Using Adaptive Image Grids6
This paper proposes an efficient bottom-up approach for detecting arbitrary-shape objects using image grids as basic components and proposes a fast adaptive grid partition method which takes image content into account and can be efficiently implemented by dynamic programming.
- 6
- 6
- 6
- 6
- 6
- 5
- Semantics-Depth-Symbiosis: Deeply Coupled Semi-Supervised Learning of Semantics and Depth5
This work tackles the MTL problem of two dense tasks, i.e., semantic segmentation and depth estimation, and presents a novel attention module called Cross-Channel Attention Module (CCAM), which facilitates effective feature sharing along each channel between the two tasks, leading to mutual performance gain with a negligible increase in trainable parameters.
- 5
- 5
- 5
- Learning location constrained pixel classifiers for image parsing5
The proposed local learning works well in different challenging image parsing problems, such as pedestrian parsing, street-view scene parsing and object segmentation, and outperforms existing results that rely on one unified pixel classifier.
- 5
- 5
- P‐79: Maximizing the 2D Viewing Field of a Computational Two‐layer Light Field 3D Display5
Experiments demonstrate that the proposed method provides a relatively desirable improvement for the whole visual effect of compressive light field display, especially for the performance in the non-target display region.
- 5
- 5
- Learning spatio-temporal dependency of local patches for complex motion segmentation5
This work presents a novel method for motion segmentation by learning the motion priors from exemplar motions to guide the segmentation, which is computationally more efficient and able to better handle the inaccurate inference of pairwise relationships.
- 5
- Fast and robust video clip search using index structure5
The mrkd-tree index structure and the proposed video representation are employed to fulfill fast and robust search of short video clips in large video collections with the total length of 15 hours.
- PoseMoE: Mixture-of-Experts Network for Monocular 3D Human Pose Estimation4
This work reveals that the depth representation is pivotal for the estimation process, and presents a Mixture-of-Experts network for monocular 3D pose estimation named PoseMoE, which outperforms the conventional lifting-based methods on three widely used datasets.
- 4
- 4
- 4
- 4
- Introduction to the Special Issue on Explainable AI on Multimedia Computing4
The paper “Precise No-Reference Image Quality Evaluation Based on Distortion Identification” by C. Yan et al. studies no-reference image quality assessment to address the challenge of lacking of knowledge about the distortion in image.
- 4
- 4
- 4
- 4
- 4
- 19‐5L: Late‐News Paper : Perceptually Optimized Dual‐layer Light Field 3D Display Using a Moiré‐aware Compressive Factorization4
Experiments demonstrate that the proposed method provides a promising perceptual optimization in both moire mitigation and definition enhancement for the visual perception of compressive light field displays.
- Body Movement Analysis and Recognition4
In this chapter, a nonverbal way of communication for human–robot interaction by understanding human upper body gestures will be addressed and an effective and real-time human gesture recognition method is proposed.
- Query-Adaptive Logo Search using Shape-Aware Descriptors4
A graph-based optimization framework to leverage category independent object proposals (candidate object regions) for logo search in a large scale image database and an efficient feature descriptor EdgeBoW, which can yield promising results, specially for object categories primarily defined by its shape.
- Aligning Instance-Semantic Sparse Representation Towards Unsupervised Object Segmentation and Shape Abstraction With Repeatable Primitives3
This work introduces a one-stage, fully unsupervised framework towards semantic-aware shape representation, providing coherent semantic interpretations of 3D object shapes across categories in a one-stage, fully unsupervised manner, without relying on annotations or heuristic semantic priors.
- Generalist Scanner Meets Specialist Locator: A Synergistic Coarse-to-Fine Framework for Robust GUI Grounding3
GMS: Generalist Scanner Meets Specialist Locator is a synergistic coarse-to-fine framework that effectively improves GUI grounding performance and significantly outperforms other strong baselines under various settings, demonstrating its robustness and potential for general-purpose GUI grounding.
- Pluralistic Salient Object Detection3
Two new SOD datasets “DUTS-MM” and “DUTS-MQ” are presented, along with newly designed evaluation metrics, and a simple yet effective pluralistic SOD baseline based on a Mixture-of-Experts (MOE) design is proposed.
- CompSlider: Compositional Slider for Disentangled Multiple-Attribute Image Generation3
This work can generate a conditional prior for the T2I foundation model to control multiple attributes simultaneously and introduces novel disentanglement and structure losses to compose multiple attribute changes while maintaining structural consistency within the image.
- 3
- A Mixed-Primitive-based Gaussian Splatting Method for Surface Reconstruction3
This paper proposes a novel framework that, for the first time, enables Gaussian Splatting to incorporate multiple types of (geometrical) primitives during its surface reconstruction process.
- FADE: A Dataset for Detecting Falling Objects Around Buildings in Video3
A novel detection method for FODB is developed that effectively leverages motion information and generates small-sized yet high-quality detection proposals and is evaluated on the proposed FADE dataset.
- 3
- 3
- OVIS: Open-Vocabulary Visual Instance Search via Visual-Semantic Aligned Representation Learning3
Through extensive experiments, ViSA's ability to search for visual instances in images not available during training given a wide range of textual queries including those composed of uncommon words is demonstrated.
- 3
- 3
- 3
- 3
- 3
- 3
- Efficient Online Spatio-Temporal Filtering for Video Event Detection3
A novel spatio-temporal filtering technique that can be implemented online and in real-time, with provable lowest computational complexity, and is compatible to any image analysis module that can produce per-pixel map of detection scores or multi-class prediction distributions.
- 3
- 3
- Unsupervised Trees for Human Action Search3
This chapter proposes a very fast action retrieval system which can effectively locate the subvolumes similar to the query video and proposes a coarse-to-fine subvolume search scheme, which results in a dramatic speedup over the existing video branch-and-bound method.
- 3
- Chain-of-Look Spatial Reasoning for Dense Surgical Instrument Counting2
Chain-of-Look, a novel visual reasoning framework that mimics the sequential human counting process by enforcing a structured visual chain, rather than relying on classic object detection which is unordered, is introduced.
- 2
- AM^2-EmoJE: Adaptive Missing-Modality Emotion Recognition in Conversation via Joint Embedding Learning2
The proposed AM^2-EmoJE not only demonstrates superior performance compared to the best-performing state-of-the-art multimodal methods, by effectively leveraging body language in place of face expression, it also exhibits an enhanced privacy feature.
- 2
- 2
- Divide and Fuse: Body Part Mesh Recovery from Partially Visible Human Images2
This work introduces a novel bottom-up approach for human body mesh reconstruction, specifically designed to address the challenges posed by partial visibility and occlusion in input images, and designs Human Part Parametric Models (HPPM) that independently reconstruct the mesh from a few shape and global-location parameters, without inter-part dependency.
- Proactive Student Persistence Prediction in MOOCs via Multi-domain Adversarial Learning2
An early “dropout” and behavior prediction model that can identify the potentially ‘marginalized’ student learning patterns to facilitate early instructional intervention in Massive Open Online Courses (MOOC) learning platform is proposed.
- 2
- Adversarial structured prediction for domain-adaptive semantic segmentation2
This work devise an adversarial structured prediction by utilizing a regularizer and makes the structured predictions follow the spatial layout learned from the source ground truth, so that the method can adapt well from a source to a target domain.
- 2
- Learning Kinematic Formulas from Multiple View Videos2
The proposed framework is trained end to end and only requires knowledge of cameras' intrinsic and extrinsic parameters and can be easily combined with the principle of parsimony, resulting in the correct free fall kinematics.
- 2
- 2
- 2
- 2
- Visual Clustering with Minimax Feature Fusion2
This chapter studies multi-feature clustering and proposes a minimax formulation to reach a consensus clustering, which can find a universal feature embedding which not only fits each feature view well, but also unifies different views by minimizing the pairwise disagreement between any two of them.
- 2
- A Compact Binary Aggregated Descriptor via Dual Selection for Visual Search2
The proposed method can effectively reduce the codesize of the raw aggregated descriptors, without degrading the search accuracy or introducing additional memory footprint, and outperforms the state-of-the-art methods.
- 2
- 2
- 2
- 1
- Tracking-seq: a universal off-target detection approach for CRISPR–Cas genome editing1
A versatile approach for assessing genome-wide off-target activity of diverse genome editors by tracking replication protein A—a key protein that binds single-stranded DNA intermediates using CUT&RUN is described.
- Vision Language Models Map Logos to Text via Semantic Entanglement in the Visual Projector1
This work investigates a previously overlooked setting: logo hallucination, where models generate brand names or textual content despite logos containing no visible words, and highlights projector disentanglement and OCR-guided decoding as promising directions for building more trustworthy multimodal systems.
- 1
- 1
- Forecasting Future Videos from Novel Views via Disentangled 3D Scene Representation1
This work proposes a disentangled two-stage approach that initially forecasts ego-motion and subsequently the residual motion of dynamic objects (e.g., cars, people) and ensures more precise motion predictions by reducing inaccuracies from entanglement of ego-motion with dynamic object motion.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- Temporally enhanced image object proposals for online video object and action detections1
A novel form of image object proposals, Temporally Enhanced Image Object Proposals (TE-IOPs), for online video object/action detection with superior performance over existing IOPs and VOPs in terms of both the proposal re-ranking and the application of online action detection.
- Visual Pattern Discovery and Recognition1
This chapter starts with an overview of the visual pattern discovery problem and then discusses the major progress of spatial and feature co-occurrence pattern discovery.
- Hierarchical Sparse Coding for Visual Co-occurrence Discovery1
This chapter investigates soft assignments instead of hard assignments used in Chap.
- 1
- Feature Co-occurrence for Visual Labeling1
This chapter proposes a transductive label propagation algorithm by leveraging the advantages of feature co-occurrence patterns in visual disambiguity by introducing a smooth regularization that ensures similar feature Co-Occurrence patterns share the same label.
- 1
- Learning a Multi-class Discriminative Dictionary with Nonredundancy Constraints for Visual Classification1
This paper concurrently exploit both the intra-class and the inter-class visual correlations to learn a multi-class discriminative dictionary and can learn a superior dictionary and an optimal linear classifier simultaneously.
- 1
- Propagative Hough Voting to Leverage Contextual Information1
The superior performances on benchmarked datasets validate that the propagative Hough voting can outperform state-of-the-art techniques in various action analysis tasks, such as action search and recognition.
- Introduction to Human Action Analysis1
This chapter starts from an overview of human action analysis problems, and then discusses the recent development of the four humanaction analysis applications as well as tree-based approaches.
- Human Action Prediction with Multiclass Balanced Random Forest1
This chapter develops a spatial-temporal implicit shape model (STISM), which characterizes the space-time structure of the sparse local features extracted from a video, and proposes a new random forest structure, called multiclass balanced random forest, which makes a good trade-off between the balance of the trees and the discriminative abilities.
- Low-Rank Online Metric Learning1
This chapter presents an online metric learning model to address the online image classification/scene recognition problem via adaptive similarity measurement and demonstrates that the low-rank property widely exists in natural data.
- Human Action Analysis with Randomized Trees1
This book describes how the randomized trees can be used for action classification, action detection, action search, and action prediction, and techniques for space-time action localization including branch-and-bound sub-volume search and propagative Hough voting.
- 1
- Rapid object search engine for contextual advertisement1
Different from the traditional image retrieval systems only matching individual points or fixed-scale spatial contexts, the proposed system considers spatial contexts of varying sizes and shapes, in the form of randomized spatial partition (RSP), and hence provides more accurate search results.
- Multimedia Data Indexing1
In this chapter,.the.authors.discuss.the challenges.of.the.users.to.make.them.accessible.in.digital.data.management.and.capabilities.
- 1
- Lens Privacy Sealing: A New Benchmark and Method for Physical Privacy-Preserving Action Recognition–
Comprehensive validation shows LPS achieves a superior privacy-utility trade-off compared to state-of-the-art hardware methods, resists reconstruction attacks including PSF inversion and data-driven recovery, and generalizes robustly across optical configurations and challenging environments.
- Ultra-sensitive profiling of CRISPR-Cas off-target effects with Tracking-seq2–
Developing Tracking-seq2, a highly sensitive method to profile off-targets for diverse genome editors, and reveal distinct off-target effects across individuals in primary human cells, which reveals that genomic variations drive distinct off-target heterogeneity across different individuals.
- De-biasing Skeleton-Based Action Recognition with Convex Hull Adaptive Shift–
This work proposes a Convex Hull Adaptive Shift based normalization method to reduce Entity bias (CHASE), improving performance across a variety of skeleton-based action and interaction recognition tasks and demonstrates compatibility with various intra-skeleton modalities, highlighting its adaptability.
- –
- Textured Geometry Evaluation: Perceptual 3D Textured Shape Metric via 3D Latent-Geometry Network–
This work proposes a new fidelity evaluation method that is based directly on 3D meshes with texture, without relying on rendering, and shows that TGE outperforms rendering-based and geometry-only methods on real-world distortion dataset.
- –
- –
- PFLIC: A novel personalized federated learning-based iterative clustering–
PFLIC, a novel personalized federated learning based on an iterative clustering algorithm, is proposed to estimate clusters to mitigate data heterogeneity and improve the efficiency of FL.
- –
- SRAM: Shape-Realism Alignment Metric for No Reference 3D Shape Evaluation–
A Shape-Realism Alignment Metric is proposed that leverages a large language model (LLM) as a bridge between mesh shape information and realism evaluation and shows that it correlates well with human perceptions and outperforms existing methods, and has good generalizability.
- –
- –
- –
- –
- A Graph-based Approach for Relating Integer Programs–
A newgraph-based representation of MILPs is proposed, where decision variables and constraints are described by nodes, and arcs denote the presence of decision variables in certain constraints, and a formal measure of graph-based structural similarity is introduced.
- –
- Real-Time Hand Pose Estimation Using Depth Camera–
Two point- set-based methods for 3D hand pose estimation from depth images are discussed: (1) point-set-based holistic regression method that directly regresses holistic 3DHand Pose estimation; and (2) Point-Set-based point-wise regression Method that generates dense outputs for robust 3Dhand pose estimation.
- –
- –
- –
- –
- –
- Automatic Visual Pattern Discovery via Cohesive Subgraph Mining–
The thematic video pattern discovery is formulated as a cohesive subgraph selection problem by finding a subset of visual words that are spatio-temporally collocated and a novel method is proposed to efficiently find the cohesive sub graph of maximum overall mutual information scores.
- Supervised Trees for Human Action Recognition and Detection–
In this chapter, a random forest is constructed to efficiently generate discriminative votes from individual interest points, and a fast top-K subvolume search algorithm is developed to find all action instances in a single round of search.
- –
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 2026-10-12. 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.