Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks506 from public data
- 980
- 860
- Person Re-Identification by Support Vector Ranking772
This work converts the person re-identification problem from an absolute scoring p roblem to a relative ranking problem and develops an novel Ensemble RankSVM to overcome the scalability limitation problem suffered by existing SVM-based ranking methods.
- 747
- 719
- 671
- 558
- Associating Groups of People510
A novel people group representation and a group matching algorithm are proposed that addresses changes in the relative positions of people in a group and the latter deals with variations in illumination and viewpoint across camera views.
- Unsupervised Person Re-Identification by Soft Multilabel Learning421
The reference agent learning to represent each reference person by a reference agent in a joint embedding is introduced and the unified deep model outperforms the state-of-the-art unsupervised RE-ID methods by clear margins.
- MIST: Multiple Instance Self-Training Framework for Video Anomaly Detection398
A multiple instance self-training framework (MIST) to efficiently refine task-specific discriminative representations with only video-level annotations, which performs comparably to or even better than existing supervised and weakly supervised methods.
- DilateFormer: Multi-Scale Dilated Transformer for Visual Recognition396
This work analyzes the patch interaction of global attention in ViTs and proposes Multi-Scale Dilated Attention (MSDA) to model local and sparse patch interaction within the sliding window, indicating the redundancy of global dependency modeling in shallow layers of ViTs.
- 359
- Cross-View Asymmetric Metric Learning for Unsupervised Person Re-Identification342
The model aims to learn an asymmetric metric, i.e., specific projection for each view, based on asymmetric clustering on cross-view person images, and finds a shared space where view-specific bias is alleviated and thus better matching performance can be achieved.
- 342
- Person Re-Identification by Camera Correlation Aware Feature Augmentation338
This work forms a novel view-specific person re-identification framework from the feature augmentation point of view, called Camera coR relation Aware Feature augmenTation (CRAFT), and presents a domain-generic deep person appearance representation which is designed particularly to be towards view invariant for facilitating cross-view adaptation by CRAFT.
- Squeeze-and-Attention Networks for Semantic Segmentation335
A novel squeeze-and-attention network (SANet) architecture is proposed that leverages an effective squeeze- and-att attention (SA) module to account for two distinctive characteristics of segmentation: i) pixel-group attention, and ii) pixels-wise prediction.
- Embedding Deep Metric for Person Re-identification: A Study Against Large Variations313
A novel moderate positive sample mining method to train robust CNN for person re-identification, dealing with the problem of large variation is proposed and the learning by a metric weight constraint is improved, so that the learned metric has a better generalization ability.
- 307
- VBench-2.0: Advancing Video Generation Benchmark Suite for Intrinsic Faithfulness299
VBench-2.0 aims to set a new standard for the next generation of video generative models in pursuit of intrinsic faithfulness, a next-generation benchmark designed to automatically evaluate video generative models for their intrinsic faithfulness.
- Person Re-Identification by Contour Sketch Under Moderate Clothing Change293
This work performs cross-clothes person re-id based on a contour sketch of person image to take advantage of the shape of the human body instead of color information for extracting features that are robust to moderate clothing change, and develops a multistream network for aggregating multi-granularity features to better identify a person.
- 286
- Top-Push Video-Based Person Re-identification261
A top-push distance learning model (TDL), in which a top- push constrain enforces the optimization on top-rank matching in re-id, is proposed to make the matching model more effective towards selecting more discriminative features to distinguish different persons.
- 244
- 238
- Fully convolutional network ensembles for white matter hyperintensities segmentation in MR images219
The effectiveness and generalization capability of the proposed system show its potential for real‐world clinical practice and are the highest achieved in the challenge, suggesting the proposed method is the state‐of‐the‐art.
- 212
- 179
- Dual Illumination Estimation for Robust Exposure Correction176
This paper presents a novel automatic exposure correction method, able to robustly produce high‐quality results for images of various exposure conditions, and demonstrates the effectiveness of the proposed approach and its superiority over the state‐of‐the‐art methods and popularautomatic exposure correction tools.
- 175
- Unsupervised Person Re-Identification by Deep Asymmetric Metric Embedding173
A novel unsupervised deep framework named the DEep Clustering-based Asymmetric MEtric Learning (DECAMEL) is developed, which learns a compact cross-view cluster structure of Re-ID data to help alleviate the view-specific bias and facilitate mining the potential cross-View discriminative information for unsuper supervised Re- ID.
- RGB-IR Person Re-identification by Cross-Modality Similarity Preservation172
This work designs a modality-gated node as a universal representation of both modality-specific and shared structures for constructing a structure-learnable feature extractor called Modality-Gated Extractor, and proposes a Focal Modality-Aware Similarity-Preserving Loss.
- Combined Depth Space based Architecture Search For Person Re-identification167
This work proposes a novel search space called Combined Depth Space (CDS), based on which it search for an efficient network architecture, which is called CDNet, via a differentiable architecture search algorithm, via the use of the combined basic building blocks in CDS.
- Quantized correlation hashing for fast cross-modal search165
This work presents a cross-modal hashing approach, called quantized correlation hashing (QCH), which takes into consideration the quantization loss over domains and the relation between domains, and outperforms the state-of-the-art multi- modal hashing methods.
- Long-Term Human Motion Prediction by Modeling Motion Context and Enhancing Motion Dynamics161
A motion context modeling by summarizing the historical human motion with respect to the current prediction is proposed, and a modified highway unit (MHU) is proposed for efficiently eliminating motionless joints and estimating next pose given the motion context.
- 159
- 159
- Shape-Erased Feature Learning for Visible-Infrared Person Re-Identification158
A shape-erased feature learning paradigm that decorrelates modality-shared features in two orthogonal subspaces achieves a conditional mutual information maximization between shape- erased feature and identity discarding body shape information, thus enhancing the diversity of the learned representation explicitly.
- 157
- 156
- Robust Depth-Based Person Re-Identification152
This work proposes to exploit depth information to provide more invariant body shape and skeleton information regardless of illumination and color change, and proposes a locally rotation invariant depth shape descriptor called Eigen-depth feature to describe pedestrian body shape.
- 141
- 133
- High-Quality Exposure Correction of Underexposed Photos130
This work addresses the problem of correcting the exposure of underexposed photos by casting the exposure correction problem as an illumination estimation optimization, where PBS is defined as three constraints for estimating illumination that can generate the desired result with even exposure, vivid color and clear textures.
- Selective Hourglass Mapping for Universal Image Restoration Based on Diffusion Model129
An advanced selective hourglass mapping strategy based on diffusion model, termed DiffUIR, which achieves state-of-the-art performance on five image restoration tasks, 22 benchmarks in the universal setting and zero-shot generalization setting and surprisingly, by only using a lightweight model, could achieve outstanding performance.
- 128
- 125
- 125
- Deep Low-Resolution Person Re-Identification124
This work develops a novel approach called Super-resolution and Identity joiNt learninG (SING) to simultaneously optimise image super- resolution and person re-id matching and introduces an adaptive fusion algorithm for accommodating multi-resolution LR images.
- Online Hashing116
This paper proposes an online hash model to accommodate data coming in stream for online learning, and extends the online hashing (OH) from a single model to a multimodel OH that trains multiple models so as to retain diverse OH models in order to avoid biased update.
- Online hashing116
This paper proposes an online hash model to accommodate data coming in stream for online learning, and extends the online hashing (OH) from a single model to a multimodel OH that trains multiple models so as to retain diverse OH models in order to avoid biased update.
- Cross-modal Consensus Network for Weakly Supervised Temporal Action Localization115
This work argues that the features extracted from the pre-trained extractors are not specific for WS-TAL task, leading to inevitable redundancy and sub-optimization, and proposes a cross-modal consensus network (CO2-Net) to tackle this problem.
- A Regularized Correntropy Framework for Robust Pattern Recognition114
A new multiple linear regression model using regularized correntropy to improve the robustness of the classical mean square error (MSE) criterion that is sensitive to outliers is proposed and a novel algorithm to solve the nonlinear optimization problem is proposed.
- 1D-LDA vs. 2D-LDA: When is vector-based linear discriminant analysis better than matrix-based?113
It is shown that 2D-LDA has eliminated the information contained in the covariance information between different local geometric structures, which is useful for discriminant feature extraction, whereas 1D- LDA could preserve such information, and this new finding indicates that 1D -LDA is able to gain higher Fisher score than 2D/LDA in some extreme case.
- Learning From Human Educational Wisdom: A Student-Centered Knowledge Distillation Method111
To address the problems of current deficiency knowledge, hard sample learning and knowledge forgetting faced by a student network in the learning process, Proportional-Integral-Derivative algorithms from automation fields are introduced and improved to make them effective in identifying the current knowledge required by the student network.
- Viewpoint-Aware Loss with Angular Regularization for Person Re-Identification106
This work proposes a novel approach to viewpoint-aware adaptive label smoothing regularization (VALSR) that assigns the adaptive soft label to feature representation and demonstrates that it outperforms the state-of-the-art supervised Re-ID methods.
- 104
- 103
- Contextual Heterogeneous Graph Network for Human-Object Interaction Detection101
A heterogeneous graph network that models humans and objects as different kinds of nodes and incorporates intra- class messages between homogeneous nodes and inter-class messages between heterogeneous nodes is proposed and exploited to improve the learning.
- Deep Bilinear Learning for RGB-D Action Recognition100
A new action feature called modality-temporal cube is presented in a tensor structure for characterizing RGB-D actions from a comprehensive perspective and the proposed method outperforms the state-of-the-art approaches.
- 98
- A Large-scale RGB-D Database for Arbitrary-view Human Action Recognition97
A large-scale RGB-D action database for arbitrary-view action analysis, including RGB videos, depth and skeleton sequences, and it is the first database containing the entire 360?
- A deep learning model and human-machine fusion for prediction of EBV-associated gastric cancer from histopathology95
A deep convolutional neural network is introduced and its fusion with pathologists for predicting Epstein–Barr virus-associated gastric cancer from histopathology and the human-machine fusion significantly improves the diagnostic performance of both the EBVNet and the pathologist.
- Real-Time RGB-D Activity Prediction by Soft Regression95
The proposed regression-based activity prediction model outperforms existing models significantly and also shows that the activity prediction on RGB-D sequence is more accurate than that on RGB channel.
- 92
- Transformer for Object Re-identification: A Survey91
A comprehensive review and in-depth analysis of the Transformer-based Re-ID is provided, categorizing existing works into Image/Video-Based Re-ID, Re-ID with limited data/annotations, Cross-Modal Re-ID, and Special Re-ID Scenarios, and thoroughly elucidate the advantages demonstrated by the Transformer.
- 89
- 88
- TW-Co-k-means: Two-level weighted collaborative k-means for multi-view clustering86
A new objective function is designed for multi-view clustering, which exploits the distinctive information in each view while taking advantage of the complementariness and consistency across different views in a collaborative manner.
- 85
- MINI-Net: Multiple Instance Ranking Network for Video Highlight Detection84
This work proposes casting weakly supervised video highlight detection modeling for a given specific event as a multiple instance ranking network (MINI-Net) learning, and forms a max-max ranking loss to acquire a reliable relative comparison between the most likely positive segment instance and the hardest negative segment instance.
- LLMDet: Learning Strong Open-Vocabulary Object Detectors under the Supervision of Large Language Models83
This work shows that an open-vocabulary detector co-training with a large language model by generating image-level detailed captions for each image can further improve performance, and shows that the improved LLMDet can in turn build a stronger large multi-modal model, achieving mutual benefits.
- 83
- 80
- 78
- Extraction of illumination invariant facial features from a single image using nonsubsampled contourlet transform78
This paper proposes to utilize the logarithmic nonsubsampled contourlet transform (LNSCT) to estimate the reflectance component from a single face image and refer it as the illumination invariant feature for face recognition, where NSCT is a fully shift-invariant, multi-scale, and multi-direction transform.
- 76
- 76
- Early detection of visual impairment in young children using a smartphone-based deep learning system73
A mobile health system, the smartphone-based Apollo Infant Sight (AIS), which identifies visually impaired children with any of 16 ophthalmic disorders by recording and analyzing their gazing behaviors and facial features under visual stimuli, has the potential to be used by healthcare professionals, parents and caregivers for identifying young children with visual impairment across a wide range of ophthalmic disorders.
- Fine-Grained Person Re-identification71
This work proposes to exploit two types of local dynamic pose features: motion-attentive local dynamic pose feature and joint-specific local dynamic pose feature that are complementary to each other and describe identity-specific pose characteristics, which are found to be more unique and discriminative against similar appearance between people.
- Learning Multi-Attention Context Graph for Group-Based Re-Identification71
A novel unified framework based on graph neural networks is proposed to simultaneously address the above two group-based re-id tasks, i.e., group re-ID and group-aware person re-ids, and a context graph with group members as its nodes to exploit dependencies among different people is constructed.
- 69
- 66
- 66
- An Asymmetric Modeling for Action Assessment65
This work proposes an asymmetric interaction module (AIM), to explicitly model asymmetric interactions between intelligent agents within an action, where they group these agents into a primary one and secondary ones (e.g., human and objects).
- Adversarial Attribute-Image Person Re-identification65
This work conducted extensive experiments on three attribute datasets and demonstrated that the regularized adversarial modelling is so far the most effective method for the attribute-image cross-modality person Re-ID problem.
- Do Not Disturb Me: Person Re-identification Under the Interference of Other Pedestrians63
A novel deep network termed Pedestrian-Interference Suppression Network (PISNet) is presented, which leverages a Query-Guided Attention Block (QGAB) to enhance the feature of the target in the gallery, under the guidance of the query.
- Lifelong Person Re-identification by Pseudo Task Knowledge Preservation61
This work casts LReID as a domain adaptation problem and proposes a pseudo task knowledge preservation framework to alleviate the task-wise domain gap, and proposes to preserve knowledge on extra pseudo tasks which jointly distills knowledge and discriminates identity.
- When Prompt-based Incremental Learning Does Not Meet Strong Pretraining60
A learnable Adaptive Prompt Generator (APG) that significantly outperforms advanced methods in exemplar-free incremental learning without (strong) pretraining and has comparable performance to existing prompt-based models, showing that the method can still benefit from pretraining.
- World-Env: Leveraging World Model as a Virtual Environment for VLA Post-Training59
World-Env is proposed, an RL-based post-training framework that replaces physical interaction with a low-cost world model-based virtual simulator that enables VLA models to safely explore and generalize beyond their initial imitation learning distribution.
- Multimodal Action Quality Assessment59
A Progressive Adaptive Multimodal Fusion Network (PAMFN) is proposed that separately models modality-specific information and mixed-modality information and argues that although AQA is highly dependent on visual information, the audio is useful complementary information for improving the score regression accuracy.
- 58
- Continual learning with Bayesian model based on a fixed pre-trained feature extractor57
Experiments reveal that the proposed Bayesian generative model for continual learning built on a fixed pre-trained feature extractor outperforms state-of-the-art approaches that even keep some images of old classes during continual learning of new classes.
- Two-Stage Sparse Representation for Robust Recognition on Large-Scale Database57
By using the two-stage sparse representation method, a significant reduction of computational cost is reached by over fifty times in comparison with the SRC, which enables the TSR to be deployed more suitably for large-scale dataset.
- Prompt-Based Learning for Unpaired Image Captioning56
This research attempts to infer the cross-domain cue information about a given image from the large VL-PTMs for the UIC task, and introduces a semantic prompt and aggregated with visual features for more accurate caption prediction under the adversarial learning framework.
- A digital mask to safeguard patient privacy56
A deep-learning algorithm that removes patient-identifying information from facial images, while retaining sufficient information for accurate disease diagnosis, has the potential to protect patient privacy and facilitate public acceptance of facial imaging for use in digital medicine.
- Multi-task mid-level feature learning for micro-expression recognition55
A novel multi-task mid-level feature learning method to enhance the discrimination ability of extracted low-level features by learning a set of class-specific feature mappings, which would be used for generating mid- level feature representation.
- 54
- Cross-Modal Adaptive Dual Association for Text-to-Image Person Retrieval53
This work demonstrates the discrepancy between image-to-text association and text-to-image association and proposescross-modal adaptive dual association (CADA) to build fine bidirectional image-text detailed associations and demonstrates the superiority of this dual formulation.
- Joint adaptive manifold and embedding learning for unsupervised feature selection53
A novel unsupervised feature selection framework, named as joint adaptive manifold and embedding learning for unsuper supervised feature selection (JAMEL) is proposed, which iteratively and adaptively learns lower-dimensional embeddings for data to preserve the manifold structure among data.
- 53
- 52
- SATS: Self-attention transfer for continual semantic segmentation50
Extensive evaluations on multiple public benchmarks support that the proposed self-attention transfer method can further effectively alleviate the catastrophic forgetting issue, and its flexible combination with one or more widely adopted strategies significantly outperforms state-of-the-art solutions.
- Abnormality Detection in Chest X-Ray Images Using Uncertainty Prediction Autoencoders50
A new abnormality detection approach based on an autoencoder which outputs not only the reconstructed normal version of the input image but also a pixel-wise uncertainty prediction, which provides a natural measurement for abnormality Detection in images.
- Adversarial Open-World Person Re-Identification48
A deep open-world group-based person re-id model based on adversarial learning to alleviate the attack problem caused by similar non-target people and is realized by the Adversarial PersonNet (APN) that jointly learns a generator, a person discriminator, a target discriminator and a feature extractor.
- 48
- Principal component analysis based on non-parametric maximum entropy48
An improved principal component analysis based on maximum entropy (MaxEnt) preservation, called MaxEnt-PCA, which is derived from a Parzen window estimation of Renyi's quadratic entropy, is proposed.
- 47
- 47
- Biometric Recognition47
A novel and effective approach to occlusion-robust face detection that combines two major phases, i.e. proposal generation and classification, and proves the competency of the proposed method at detecting occluded faces.
- 47
- 46
- Large-Scale Visible Watermark Detection and Removal with Deep Convolutional Networks46
This work proposes the first general deep learning based framework, which can precisely detect and remove a variety of watermark with convolutional networks, and aims to increase copyright awareness for the spread of online images.
- 46
- 45
- 45
- Weakly Supervised Person Re-Identification44
A Cross-View MIML (CV-MIML) method is developed that is able to explore potential intraclass person images from all the camera views by incorporating the intra-bag alignment and the cross-view bag alignment.
- 44
- ViSpeak: Visual Instruction Feedback in Streaming Videos43
The ViSpeak model is proposed, which is a SOTA streaming video understanding LMM with GPT-4o-level performance on various streaming video understanding benchmarks and is equipped with basic visual instruction feedback ability, serving as a solid baseline for future research.
- 43
- 42
- 42
- Action Knowledge Transfer for Action Prediction with Partial Videos42
The proposed action knowledge transfer method can significantly improve the performance of action prediction, especially for the actions with small observation ratios (e.g., 10%).
- Representation Learning for Scene Graph Completion via Jointly Structural and Visual Embedding42
Experimental results demonstrate that the proposed RLSV model outperforms all baselines in both tasks, which justifies the significance of combining structural and visual information for scene graph completion.
- 42
- 41
- S3D-CNN: skeleton-based 3D consecutive-low-pooling neural network for fall detection41
A skeleton-based 3D consecutive-low-pooling neural network (S3D-CNN) for fall detection is proposed, which is evaluated on public and self-collected datasets respectively, outperforming the existing methods.
- Towards Photo-Realistic Visible Watermark Removal with Conditional Generative Adversarial Networks41
A new watermark processing framework using the conditional generative adversarial networks (cGANs) for visible watermark removal in the real-world application is presented and it is shown that the proposed approach can produce more photo-realistic and convincing results compared with the state-of-the-art methods.
- How many clusters? A robust PSO-based local density model41
This paper aims to develop a novel local and not differentiable clustering method based on Particle Swarm Optimization, which can estimate the number of clusters automatically, and measures the local compactness of each cluster by local density function, pushes the PSO towards maximizing such a compactness.
- PCCT: Progressive Class-Center Triplet Loss for Imbalanced Medical Image Classification40
Extensive experiments support that the PCCT framework works effectively for medical image classification with imbalanced training images, achieving state-of-the-art performance and outperforming the widely used methods for the class imbalance issue.
- 40
- ReferDINO: Referring Video Object Segmentation with Visual Grounding Foundations39
This work proposes ReferDINO, a strong RVOS model that inherits region-level vision-language alignment from foundational visual grounding models, and is further endowed with pixellevel dense perception and cross-modal spatiotemporal reasoning.
- Ultra-wide FOV meta-camera with transformer-neural-network color imaging methodology38
This work designs a neural meta-camera by introducing a knowledge-fused data-driven paradigm equipped with transformer-based network, which enables the network to sequentially assimilate the physical prior and experimental data of the metalens, and thus can effectively mitigate the aforementioned challenges of wide FOV.
- Learning to Imagine: Diversify Memory for Incremental Learning using Unlabeled Data38
This work develops a learnable feature generator to diversify exemplars by adaptively generating diverse counterparts of exemplars based on semantic information from exemplars and semantically-irrelevant information from unlabeled data.
- Illumination invariant single face image recognition under heterogeneous lighting condition38
A novel gradient based descriptor, namely Logarithm Gradient Histogram (LGH), which takes the illumination direction, magnitude and the spectral wavelength together into consideration, so that it can handle both homogeneous and heterogeneous lightings.
- Class attention to regions of lesion for imbalanced medical image recognition37
A simple yet effective framework, named CARE, to handle data imbalance issues by embedding attention into the training process of CNNs and helps CNNs attend to lesion regions of rare diseases, therefore helping CNNs to learn their characteristics more effectively.
- DBDNet: Learning Bi-directional Dynamics for Early Action Prediction37
A novel encoder-decoder architecture is proposed for integrating the tasks of synthesizing future motions from observed videos and reconstructing observed motions from synthesized future motions in an unified framework, which can capture the bi-directional dynamics depicted in partial videos along the temporal (past-to-future) direction and reverse chronological direction.
- Decoupled Distillation to Erase: A General Unlearning Method for Any Class-centric Tasks36
This work proposes a theoretical framework to analyze the general form of unlearning loss and decompose it into forgetting and retention terms, and proposes a mask distillation unlearning method that optimizes both the forgetting and refined retention components simultaneously.
- Deep kNN for Medical Image Classification36
An end-to-end learning strategy to unify the kNN classification and the feature extraction procedure and shows that the proposed deep kNN outperforms both kNN and other strong classifiers.
- Spatial–temporal consistent labeling of tracked pedestrians across non-overlapping camera views36
A Bayesian model to solve the consistent labeling problem across multiple non-overlapping camera views and an online method for online correspondence update using optimal graph matching (OGM) algorithm is developed.
- PSLT: A Light-Weight Vision Transformer With Ladder Self-Attention and Progressive Shift35
A ladder self-attention block with multiple branches and a progressive shift mechanism to develop a light-weight transformer backbone that requires less computing resources (e.g., a relatively small number of parameters and FLOPs), termed Progressive Shift Ladder Transformer (PSLT).
- Cross-View Relation Networks for Mammogram Mass Detection35
A novel mass detection framework to capture the latent relation information from the two paired views of a same mass in mammogram is proposed, demonstrating that the proposed method outperforms existing feature fusion approaches and state-of-the-art mass detection methods.
- 35
- 34
- Continual Learning of New Diseases with Dual Distillation and Ensemble Strategy34
A new continual learning framework is proposed to alleviate the issue of catastrophic forgetting of old knowledge when learning new classes by simultaneously distilling both old knowledge and recently learned new knowledge and by ensembling the class-specific knowledge from the previous classifier and the learned new classifier.
- Predicting Future Instance Segmentation with Contextual Pyramid ConvLSTMs34
A novel framework is proposed, which contains a set of ConvLSTMs to exploit intra-level spatio-temporal contexts for predicting features of each individual level, and which can achieve state-of-the-art performance on two video instance segmentation benchmarks.
- Estimator Meets Equilibrium Perspective: A Rectified Straight Through Estimator for Binary Neural Networks Training33
A new perspective to the BNNs training is presented, regarding it as the equilibrium between the estimating error and the gradient stability, and a power function based estimator, Rectified Straight Through Estimator (ReSTE for short), which has excellent performance and surpasses the state-of-the-art methods without any auxiliary modules or losses.
- 33
- 33
- Deep Dual Relation Modeling for Egocentric Interaction Recognition33
A novel interactive LSTM module is developed to explicitly model the relations between the two interacting persons based on their individual action representations, which are collaboratively learned with an interactor attention module and a global-local motion module.
- 33
- 33
- 32
- 32
- Learning to Learn Relation for Important People Detection in Still Images32
This work proposes a deep imPOrtance relatIon NeTwork (POINT) that combines both relation modeling and feature learning and demonstrates the efficacy of learning to learn relations for important people detection.
- A Versatile Framework for Multi-Scene Person Re-Identification31
This work contributes to the first attempt at learning a versatile ReID model that can handle various ReID challenges at the same time, and introduces the multi-scene properties into self-supervised learning of ReID via a multi-scene prioris data augmentation (MPDA) strategy.
- Cross-Camera Feature Prediction for Intra-Camera Supervised Person Re-identification across Distant Scenes31
A cross-camera feature prediction method to mine cross- Camera self supervision information from camera-specific feature distribution by transforming fake cross- camera positive feature pairs and minimize the distances of the fake pairs is proposed.
- 31
- Deep Camouflage Images31
A novel neural style transfer approach that adopts the visual perception mechanism to create camouflage images, which allows us to hide objects more effectively while producing natural-looking results.
- Multi-view collaborative locally adaptive clustering with Minkowski metric31
A novel approach termed multi-view collaborative locally adaptive clustering with Minkowski metric (MV-CoMLAC) is proposed, capable of simultaneously taking into account the subspace diversity within each view as well as the knowledge across different views.
- Joint Bilateral-Resolution Identity Modeling for Cross-Resolution Person Re-Identification30
An important discovery is that leveraging LR-specific identity information enables a simple cascade of super-resolution and person re-id learning to achieve state-of-the-art performance, without elaborate model design nor bells and whistles, which has not been investigated before.
- 29
- Weakly Supervised Discriminative Feature Learning With State Information for Person Identification29
A simple pseudo label model is formulated and the state information is utilized as weak supervision in an attempt to refine the assigned pseudo labels by the weakly supervised decision boundary rectification andWeakly supervised feature drift regularization.
- CARE: Class Attention to Regions of Lesion for Classification on Imbalanced Data29
Comprehensive experiments showed that paying attention to lesion regions of rare diseases during learning not only improved the classification performance on rare diseases, but also on the mean class accuracy.
- Maximal granularity structure and generalized multi-view discriminant analysis for person re-identification29
The proposed descriptor of MGSD captures rich local structural information from overlapping macro-pixels in an image, analyzes the horizontal occurrence of multi-granularity and maximizes the occurrence to extract a robust representation for viewpoint changes, and can obtain rich person appearance whilst being robust against different condition changes.
- 28
- One-pass person re-identification by sketch online discriminant analysis28
This work develops an Sketch online Discriminant Analysis by embedding sketch processing into Fisher discriminant analysis (FDA), which can efficiently keep the main data variations of all passed samples in a low rank matrix when processing sequential data samples, and provides theoretical analysis on the effect of the estimated approximate within-class covariance matrix.
- 28
- 27
- Rethinking Bimanual Robotic Manipulation: Learning with Decoupled Interaction Framework27
A novel decoupled interaction framework that considers the characteristics of different tasks in bimanual manipulation, to assign an independent model to each arm to enhance the learning of uncoordinated tasks, while introducing a selective interaction module that adaptively learns weights from its own arm to improve the learning of coordinated tasks.
- Deep asymmetric video-based person re-identification27
A Deep Asymmetric Metric learning (DAM) method is proposed that embeds a proposed asymmetric distance metric learning loss into a two-stream deep neural network for jointly learning view-specific and feature-specific transformations to overcome the “view-bias” problem in video-based person re-id.
- Improving Robustness of Medical Image Diagnosis with Denoising Convolutional Neural Networks27
This paper proposes a novel mechanism to improve the robustness of medical image classification systems by bringing denoising ability to CNN classifiers with a naturally embedded auto-encoder and high-level feature invariance to general noises.
- 27
- DiffuVolume: Diffusion Model for Volume based Stereo Matching26
This work pioneeringly applies the diffusion model to stereo matching to construct a more precise cost volume and outperforms all the published methods on Scene Flow, KITTI2012, KITTI2015 benchmarks and zero-shot generalization setting.
- 26
- Pyramid Texture Filtering26
This work shows that the approach is effective to separate structure from texture of different scales, local contrasts, and forms, without degrading structures or introducing visual artifacts, and demonstrates the applicability of the method on various applications including detail enhancement, image abstraction, HDR tone mapping, inverse halftoning, and LDR image enhancement.
- AC2AS: Activation Consistency Coupled ANN-SNN framework for fast and memory-efficient SNN training26
A SNN-to-ANN (SNN2ANN) framework to train the SNN in a fast and memory-efficient way and an adaptive threshold adjustment (ATA) algorithm to address the noisy spike problem is proposed.
- Multiple metric learning based on bar-shape descriptor for person re-identification26
This paper proposes a new Multiple Bar-shape Descriptor that can take full account of the spatial correlation between the center points and their adjacent points on different directions and provides an effective multiple metric learning method fusing two-channel bar-shape structural features via the optimal similarity pairwise measure obtained by a dissimilarity matrix.
- Case-based teaching using the Laboratory Animal System for learning C/C++ programming26
Students can learn C/C++ programming and gain knowledge more quickly and effectively using the case teaching model than through some traditional methods of teaching.
- 25
- Event-Guided Procedure Planning from Instructional Videos with Text Supervision25
An Event-guided Prompting-based Procedure Planning (E3P) model is contributed, which encodes event information into the sequential modeling process to support procedure planning and adopts a mask-and-predict approach for relation mining, incorporating a probabilistic masking scheme for regularization.
- Discriminant subspace learning constrained by locally statistical uncorrelation for face recognition25
A new linear dimension reduction method called locally uncorrelated discriminant projections (LUDP) is proposed, which addresses the two problems from a new aspect, and aims to decorrelate learned discriminant factors over data locally rather than globally.
- Perturbation LDA: Learning the difference between the class empirical mean and its expectation25
P-LDA is developed, in which perturbation random vectors are introduced to learn the effect of the difference between the class empirical mean and its expectation in Fisher criterion, and which outperforms the popular Fisher's LDA-based algorithms in the undersampled case.
- ChainHOI: Joint-based Kinematic Chain Modeling for Human-Object Interaction Generation24
ChainHOI introduces a novel joint graph to capture potential interactions with objects, and a Generative Spatiotemporal Graph Convolution Network to explicitly model interactions at the joint level, ensuring more realistic and biomechanically coherent motions.
- 24
- 24
- 24
- SIOD: Single Instance Annotated Per Category Per Image for Object Detection24
Degraded from inter-task (WSOD) or inter-image (SSOD) discrepancies to the intra-image discrepancy, SIOD provides more reliable and rich prior knowledge for mining the rest of unlabeled instances and trades off the annotation cost and performance.
- Weakly Supervised Open-Set Domain Adaptation by Dual-Domain Collaboration24
The Collaborative Distribution Alignment method is proposed, which performs knowledge transfer bilaterally and works collaboratively to classify unlabeled data and identify outlier samples and achieves state-of-the-art performance.
- 23
- Learning Heterogeneous Dictionary Pair with Feature Projection Matrix for Pedestrian Video Retrieval via Single Query Image23
A joint feature projection matrix and heterogeneous dictionary pair learning (PHDL) approach for IVPR is proposed and to ensure that the obtained coding coefficients have favorable discriminability, PHDL designs a point-to-set coefficient discriminant term.
- Sculpting Holistic 3D Representation in Contrastive Language-Image-3D Pre-Training22
MixCon3D, a simple yet effective method aiming to sculpt holistic 3D representation in contrastive language-image-3D pre-training, surpassing the previous state-of-the-art performance on the challenging 1,156- category Objaverse-LVIS dataset by 5.7%.
- Continual Action Assessment via Task-Consistent Score-Discriminative Feature Distribution Modeling22
The idea for modeling Continual-AQA is to sequentially learn a task-consistent score-discriminative feature distribution, in which the latent features express a strong correlation with the score labels regardless of the task or action types.
- Rethinking Temporal Fusion for Video-Based Person Re-Identification on Semantic and Time Aspect22
A novel general temporal fusion framework to aggregate frame features on both semantic aspect and time aspect and a multi-stage fusion network is explored to fuse richer frame features at multiple semantic levels, which can effectively reduce the information loss caused by the traditional single- stage fusion.
- Unsupervised Learning for Optical Flow Estimation Using Pyramid Convolution LSTM22
This paper proposes an unsupervised optical flow estimation framework named PCLNet, which uses pyramid Convolution LSTM (ConvLSTM) with the constraint of adjacent frame reconstruction, which allows flexibly estimating multi-frame optical flows from any video clip.
- PersonRank: Detecting Important People in Images22
A multiple HyperInteraction Graph that treats each individual in an image as a node and inferring the most active node from the interactions estimated by using various types of cues is constructed, and the proposed PersonRank outperforms related methods clearly and substantially.
- 22
- One-pass online learning: A local approach22
A local online learning (LOL) method, a multiple hyperplane Passive Aggressive algorithm integrated with online clustering, so that all local hyperplanes are learned jointly and working cooperatively by formulating a common component as information traffic among multiple hyperplanes in LOL.
- 21
- Human-Centric Transformer for Domain Adaptive Action Recognition21
The proposed Human-Centric Transformer develops a decoupled human-centric learning paradigm to explicitly concentrate on human-centric action cues in domain-variant video feature learning and exploits domain-invariant interaction between humans and action-correlated contexts.
- 21
- A Multi-Level Relation-Aware Transformer model for occluded person re-identification21
A Multi-Level Relation-Aware Transformer (MLRAT) model for occluded person Re-ID is introduced, which significantly outperforms existing baselines on four occluded datasets, while maintains top performance on two partial datasets and two holistic datasets.
- Automatic Modelling for Interactive Action Assessment21
A framework for modelling asymmetric interactions among agents for action assessment, considering the subordinations among agents in many interactive actions is formulated and the extensive experiments on the AQA-7 dataset indicate the robustness of the model in conventional action assessment settings.
- 21
- Regularized Locality Preserving Learning of Pre-Image Problem in Kernel Principal Component Analysis21
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment20
A novel Action-Transition-Aware Boundary Alignment (ATBA) framework is proposed to efficiently and effectively filter out noisy boundaries and detect transitions and to boost the semantic learning in the case that noise is inevitably present in the pseudo segmentation.
- 20
- 20
- One for More: Selecting Generalizable Samples for Generalizable ReID Model20
A one-for-more training objective that directly takes the generalization ability of selected samples as a loss function and learn a sampler to automatically select generalizable samples that can effectively improve the ReID model training and boost the performance of ReID models.
- Biomarker Localization by Combining CNN Classifier and Generative Adversarial Network20
A novel deep neural network architecture to effectively localize potential biomarkers in medical images, when only the image-level labels are available during model training, showing state-of-the-art performance in localizing biomarkers even if biomarkers are irregularly scattered and are of various sizes in images.
- Discrimination of the behavioural dynamics of visually impaired infants via deep learning20
The analysis of behavioural patterns from standardized video recordings of infants with varying degrees of visual impairment enables, via deep learning, classification of the infants by visual-impairment severity and by ophthalmological condition.
- Modeling Multiple Normal Action Representations for Error Detection in Procedural Tasks19
An Adaptive Multiple Normal Action Representation (AMNAR) framework that predicts all valid next actions and reconstructs their corresponding normal action representations, which are compared against the ongoing action to detect errors.
- Classifier-Head Informed Feature Masking and Prototype-Based Logit Smoothing for Out-of-Distribution Detection19
This study proposes an effective post-hoc OOD detection method, named HIMPLoS, based on a new feature masking strategy and a novel logit smoothing strategy that can reduce feature activation of OOD data and enlarge the gap in OOD score between ID and OOD data.
- 19
- Unpaired Image Captioning by Image-Level Weakly-Supervised Visual Concept Recognition19
This is the first attempt to address the problem of Weakly-Supervised visual concept recognition for UIC (WS-UIC) based only on image-level labels and design an unrecognized object (UnO) loss to improve the alignment of the inferred object and relationship information with the images.
- PC-Dance: Posture-controllable Music-driven Dance Synthesis19
A powerful framework named PC-Dance is proposed to perform adaptive posture-controllable music-driven dance synthesis and an efficient scheme for adaptive motion graph construction (AMGC) is introduced, which could improve the efficiency of graph-based optimization and preserve the diversity of motions.
- 19
- 19
- Latent embeddings for collective activity recognition19
The embeddings of latent variables build a global relation containing person-group interactions and richer contextual information by jointly modeling broader range of individuals by embedding latent variables into feature space and learning the feature mapping functions in a deep learning framework.
- Extracting non-negative basis images using pixel dispersion penalty19
A pixel dispersion penalty (PDP) is proposed, which effectively describes the spatial dispersion of pixels in an image without using any manually predefined structured patterns as constraints, and developed a spatial non-negative matrix factorization (Spatial NMF) and a spatialnon-negative component analysis (Sp spatial NCA).
- 18
- 18
- 18
- 18
- Letter-Level Online Writer Identification18
This work addresses the variance of online writing styles (Var-O-Styles) in a capture-normalize-aggregate fashion, and contributes a large-scale LEtter-level online wRiter IDentification dataset (LERID) for evaluation.
- Towards Unbiased Covid-19 Lesion Localisation And Segmentation Via Weakly Supervised Learning18
A data-driven framework that can explicitly separate potential lesions from original images, with the help of an generative adversarial network and a lesion-specific decoder is proposed.
- Learning to Detect Important People in Unlabelled Images for Semi-Supervised Important People Detection18
This work iteratively learns to assign pseudo-labels to individuals in un-annotated images and learns to update the important people detection model based on data with both labels and pseudo-label estimation, and introduces a ranking strategy for pseudo- label estimation.
- Robust joint representation with triple local feature for face recognition with single sample per person18
A novel model, named robust joint representation with triple local feature (RJR-TLF), to address this issue from the viewpoints of feature extraction and classifier design, and is much more robust to various facial variations than the recent FR with SSPP methods.
- Panorama Generation From NFoV Image Done Right17
PanoDecouple is proposed, a decoupled diffusion model framework, which decouples the panorama generation into distortion guidance and content completion, aiming to generate panoramas with both accurate distortion and visual appeal.
- 17
- 17
- 16
- 16
- 16
- Adversarial Partial Domain Adaptation by Cycle Inconsistency16
This work proposes to filter out source samples of outlier classes by weight suppression and align the distributions of shared classes between the source and target domains by adversarial learning and designs cycle transformations based on domain prototypes and soft nearest neighbor.
- 16
- A Decomposition Algorithm for the Sparse Generalized Eigenvalue Problem16
A new effective decomposition method that uses random or/and swapping strategies to find a working set and perform global combinatorial search over the small subset of variables to solve the sparse generalized eigenvalue problem.
- 16
- 16
- 15
- RoGSplat: Learning Robust Generalizable Human Gaussian Splatting from Sparse Multi-View Images15
RoGSplat is presented, a novel approach for synthesizing high-fidelity novel views of unseen human from sparse multi-view images, while requiring no cumbersome per-subject optimization and outperforms state-of-the-art methods in novel view synthesis and cross-dataset generalization.
- Single-View Scene Point Cloud Human Grasp Generation15
This work introduces S2HGrasp, a framework composed of two key modules: the Global Perception module that globally perceives partial object point clouds and the DiffuGrasp module designed to generate high-quality human grasps based on complex inputs that include scene points.
- Pedestrian re-Identification Based on Tree Branch Network with Local and Global Learning15
Experimental results demonstrate the effectiveness of the global and local feature learning method in the proposed TBN framework and show significant improvement in performance over state-of-the-art methods on three public benchmarks: Market-1501, CUHK-03 and DukeMTMC.
- Ensemble Convolutional Neural Networks for Cell Classification in Microscopic Images15
This challenge addresses the cell classification problem using deep convolutional neural networks (CNNs) and an ensemble of state-of-the-art CNNs was adopted, and the Grad-CAM technique was used to check the stability of the CNN model.
- 15
- Loc4Plan: Locating Before Planning for Outdoor Vision and Language Navigation14
A novel framework, Locating before Planning (Loc4Plan), designed to incorporate spatial perception for action planning in outdoor VLN tasks, and extensive experiments show that the proposed Loc4Plan outperforms the SOTA methods.
- FeatWalk: Enhancing Few-Shot Classification through Local View Leveraging14
This paper proposes a simple yet effective few-shot learning method called FeatWalk which can utilize the complementary nature of global and local views, therefore providing an intuitive and effective solution to the problem of insufficient local information extraction from the pre-trained feature extractor.
- ActionHub: A Large-scale Action Video Description Dataset for Zero-shot Action Recognition14
A novel Cross-modality and Cross-action Modeling (CoCo) framework for ZSAR that significantly outperforms the state-of-the-art on three popular ZSAR benchmarks (i.e., Kinetics-ZSAR, UCF101 and HMDB51) under two different learning protocols in ZSAR.
- 14
- 14
- 14
- Distilled Camera-Aware Self Training for Semi-Supervised Person Re-Identification14
A Distilled Camera-Aware Self Training framework for semi-supervised person re-identification, and a Multi-Teacher Selective Similarity Distillation Loss to selectively aggregate the knowledge of multiple weak teacher models trained with different subsets and distill a stronger student model for self training.
- 14
- Group Association: Assisting Re-identification by Visual Context14
A novel people group representation and a group matching algorithm are proposed that addresses changes in the relative positions of people in a group and the latter uses the proposed group descriptors for measuring the similarity between two candidate images.
- A Survey of Face Hallucination14
The strengths and weaknesses of approaches to high quality face hallucination are identified to form a base for a new sparsity-based method for super-resolving mis-aligned face images.
- TypeTele: Releasing Dexterity in Teleoperation by Dexterous Manipulation Types13
TypeTele, a type-guided dexterous teleoperation system, which enables dexterous hands to perform actions that are not constrained by human motion patterns by introducing dexterous manipulation types into the teleoperation system, allowing operators to employ appropriate types to complete specific tasks.
- DNF-Intrinsic: Deterministic Noise-Free Diffusion for Indoor Inverse Rendering13
DNF-Intrinsic is presented, a robust yet efficient inverse rendering approach fine-tuned from a pre-trained diffusion model, where it is proposed to take the source image rather than Gaussian noise as input to directly predict deterministic intrinsic properties via flow matching.
- Task-Oriented 6-DoF Grasp Pose Detection in Clutters13
This paper investigates the problem named Task-Oriented 6-DoF Grasp Pose Detection in Clutters (TO6DGC), which extends the task-oriented problem to a more general 6-DOF Grasp Pose Detection in Cluttered (multi-object) scenario, and proposes One-Stage TaskGrasp (OSTG), a strong baseline to address the TO6DGC problem.
- A Hierarchical Semantic Distillation Framework for Open-Vocabulary Object Detection13
This work proposes a hierarchical semantic distillation framework named HD-OVD to construct a comprehensive distillation process, which exploits generalizable knowledge from the CLIP model in three aspects and inherits generalizable recognition ability from CLIP in instance, class, and image levels.
- 13
- Siamese Learning with Joint Alignment and Regression for Weakly-Supervised Video Paragraph Grounding13
This work proposes a novel siamese learning framework that jointly learns the cross-modal feature alignment and temporal coordinate regression without timestamp labels to achieve concise one-stage localization for Weakly-Supervised Video Paragraph Grounding.
- Factorized Diffusion Autoencoder for Unsupervised Disentangled Representation Learning13
A content decorrelation loss and a mask entropy loss to decorrelate content masks in latent space and spatial space are proposed and show that the method achieves advanced performance and can generate visually interpretable concept-specific masks.
- Learning to Remove Shadows from a Single Image13
Self-ShadowGAN is presented, a novel adversarial framework that is able to learn to remove shadows in an image by training solely on the image itself, using the shadow mask as the only supervision.
- 13
- 13
- A Block Decomposition Algorithm for Sparse Optimization13
A new block decomposition algorithm that combines the effectiveness of combinatorial search methods and the efficiency of coordinate descent methods, and generally outperforms the well-known greedy pursuit method is considered.
- MIXGAN: Learning Concepts from Different Domains for Mixture Generation13
A mixture generative adversarial network (MIXGAN) is proposed, which learns concepts of content and style from two domains respectively, and thus can join them for mixture generation in a new domain, i.e., generating images with content from one domain andstyle from another.
- 13
- Sparse transfer for facial shape-from-shading13
An image-based 3D face shape reconstruction method which transfers shape cues inferred from source face images to guide the reconstruction of the target face and can produce visually better facial details and obtain smaller reconstruction errors.
- A facial sparse descriptor for single image based face recognition13
A new framework for detecting feature keypoints sparsely, describing feature context and matching feature points between two face images is developed, which is called Facial Sparse Descriptor (FSD).
- WeDetect: Fast Open-Vocabulary Object Detection as Retrieval12
This work shows that, with well-curated data and full training, the non-fusion WeDetect surpasses other fusion models and establishes a strong open-vocabulary foundation, and proposes WeDetect-Ref, an LMM-based object classifier to handle complex referring expressions, which retrieves target objects from the proposal list extracted by WeDetect-Uni.
- FodFoM: Fake Outlier Data by Foundation Models Creates Stronger Visual Out-of-Distribution Detector12
A novel OOD detection framework FodFoM is proposed that innovatively combines multiple foundation models to generate two types of challenging fake outlier images for classifier training and shows that image classifiers with the help of constructed fake images can more accurately differentiate real OOD image from ID ones.
- 12
- 12
- Weakly-Supervised Temporal Action Localization by Progressive Complementary Learning12
This work addresses the Weakly-Supervised Temporal Action Localization from a novel category exclusion perspective, which gradually enhances the snippet-level supervision to bridge the gap between video-level supervision and unavailable snippet-level supervision.
- 12
- 12
- Joint regression and learning from pairwise rankings for personalized image aesthetic assessment12
A novel approach for predicting personalized image aesthetics that fit an individual user’s personal taste by joint regression and learning from pairwise rankings, which is clearly outperforming state-of-the-art methods.
- 12
- ActionArt: Advancing Multimodal Large Models for Fine-Grained Human-Centric Video Understanding11
This work introduces ActionArt, a fine-grained video-caption dataset designed to advance research in human-centric multimodal understanding, and proposes proxy tasks to enhance the model perception ability in both spatial and temporal dimensions.
- 11
- Continual learning for seizure prediction via memory projection strategy11
Memory Projection can operate with minimal storage and computational cost, which demonstrates practical potential for seizure prediction in clinical scenarios and implements a continual learning CL strategy Memory Projection for epilepsy prediction.
- 11
- Data Augmentation is More Important Than Model Architectures for Retinal Vessel Segmentation11
By simply augmenting data with oriented image patches extracted from the limited training images, even a very simple U-Net with these augmented training patches can outperform the state-of-the-art models with much more complicated architectures or training schemes.
- CycleManip: Enabling Cyclic Task Manipulation via Effective Historical Perception and Understanding10
The CycleManip framework is proposed to achieve cycle-based task manipulation in an end-to-end imitation manner without requiring any extra models, hierarchical structure or significant computational overhead, and to enhance effective history perception by a cost-aware sampling strategy and to improve historical understanding by multi-task learning.
- 10
- 10
- 10
- Progressive Human Motion Generation Based on Text and Few Motion Frames10
A novel Progressive Motion Generation (PMG) method to progressively generate a motion from the frames with low uncertainty to those with high uncertainty in multiple stages, outperforms existing T2M generation methods by a large margin with even one given frame.
- 10
- 10
- 10
- PixelFade: Privacy-preserving Person Re-identification with Noise-guided Progressive Replacement10
This paper proposes Noise-guided Objective Function with the feature constraints of a specific authorization model, optimizing pedestrian images to normal-distributed noise images while preserving their original identity information as per the authorization model.
- 10
- 10
- 10
- 10
- 9
- TagFog: Textual Anchor Guidance and Fake Outlier Generation for Visual Out-of-Distribution Detection9
A new learning framework is proposed which leverage simple Jigsaw-based fake OOD data and rich semantic embeddings from the ChatGPT description of ID knowledge to help guide the training of the image encoder.
- Retinal Artery/Vein Classification via Rotation Augmentation and Deeply Supervised U-net Segmentation9
This study uses one stage multiclass segmentation without any graph-based or vote-based post processing to solve the artery/vein classification problem directly and effectively and shows that artery-vein-background segmentation based on deep learning can be used as a promising method for arteriovenous classification and can be combined with conventional methods for better results.
- 9
- Ensemble Transductive Learning for Skin Lesion Segmentation9
A transductive segmentation approach by choosing some of the pixels in test images to participate the training of any segmentation model together with the training set is proposed, so that visual features in the test images can be effectively learned during model training.
- One-Snapshot Face Anti-spoofing Using a Light Field Camera9
A new approach to utilizing light field camera for defending spoofing face attacks, like (warped) printed 2D facial photos and high-definition tablet images, can achieve remarkable anti-spoofing accuracy under different types of spoofing attacks.
- Robust spectral regression for face recognition9
This paper studies robust spectral regression based discriminant subspace learning from correntropy and spatially smooth structure of facial subspace through a compound regularization model that can be efficiently optimized via an accelerated proximal gradient algorithm.
- Long-RVOS: A Comprehensive Benchmark for Long-term Referring Video Object Segmentation8
ReferMo is proposed, a promising baseline method that integrates motion information to expand the temporal receptive field, and employs a local-to-global architecture to capture both short-term dynamics and long-term dependencies.
- Ranking Distillation for Open-Ended Video Question Answering with Insufficient Labels8
This work introduces a simple yet effective ranking distillation framework (RADI) that employs a teacher model trained with incomplete labels to generate rankings for potential answers, which contain rich knowledge about label priority as well as label-associated visual cues, thereby enriching the insufficient labeling information.
- ASAG: Building Strong One-Decoder-Layer Sparse Detectors via Adaptive Sparse Anchor Generation8
This work proposes Adaptive Sparse Anchor Generator (ASAG) which predicts dynamic anchors on patches rather than grids in a sparse way so that it alleviates the feature conflict problem and achieves a better speed-accuracy trade-off.
- 8
- 8
- Approximate kernel competitive learning8
A pseudo-parallelled approximatekernel competitive learning (PAKCL) based on a set-based kernel competitive learning strategy is proposed, which overcomes the obstacle of using parallel programming in Kernel competitive learning and significantly accelerates the approximate kernel competitiveLearning for large scale clustering.
- 7
- NECA: Neural Customizable Human Avatar7
This work introduces NECA, an approach capable of learning versatile human representation from monocular or sparse-view videos, enabling granular customization across aspects such as pose, shadow, shape, lighting and texture.
- 7
- Real-to-Sim Grasp: Rethinking the Gap between Simulation and Real World in Grasp Detection7
This work proposes a Real-to-Sim framework for 6-DoF Grasp detection, named R2SGrasp, with the key insight of bridging this gap in a real-to-sim way, which directly bypasses the camera noise in grasp detector training through an inference-time real-to-sim adaption.
- Cross-Camera Trajectories Help Person Retrieval in a Camera Network7
A novel cross-camera spatio-temporal model that integrates pedestrians’ walking habits and the path layout between cameras to form a joint probability distribution is proposed, which can be specified using sparsely sampled pedestrian data.
- 7
- AR-CNN: an attention ranking network for learning urban perception7
This work forms an end-to-end learning methodology to process input images according to color and texture information before inputting it into the neural network in the AR-CNN model.
- Joint graph regularized dictionary learning and sparse ranking for multi-modal multi-shot person re-identification7
This paper proposes a novel joint graph regularized dictionary learning and sparse ranking method for multi-modal multi-shot person Re-ID and designs an APG based optimization to solve the problem.
- Transductive Multi-Object Tracking in Complex Events by Interactive Self-Training7
A transductive interactive self-training method to adapt the tracking model to unseen crowded scenes with unlabeled testing data by means of teacher-student interative learning and to reduce prediction variance in an unseen domain is proposed.
- Efficient and Switchable CNN for Crowd Counting Based on Embedded Terminal7
A tiny model based on convolutional neural networks is proposed for crowd counting by edge computing in embedded terminals that supports variant sizes of input images, which benefits the applications to a variety of embedded devices.
- Learning object-specific DAGs for multi-label material recognition7
A multi-label learning framework for identifying multiple materials of a real-world object surface without a segmentation for each of them and finds that there are potential correlations between materials and that correlations are relevant to object category.
- 7
- RADICAL EXTRACTION USING AFFINE SPARSE MATRIX FACTORIZATION FOR PRINTED CHINESE CHARACTERS RECOGNITION7
A radical-based Chinese character recognition model is developed, which performs a far smaller category classification than the whole character-based recognition, resulting in a more robust recognition system.
- 7
- Non-ideal class non-point light source quotient image for face relighting7
The proposed NIC-NPL-QI algorithm obtains better quality in synthesizing face images as compared with state-of-the-art algorithms and can handle the harmonic light and shadows.
- 6
- 6
- 6
- 6
- 6
- 6
- 6
- STVGFormer: Spatio-Temporal Video Grounding with Static-Dynamic Cross-Modal Understanding6
A concise and effective framework named STVGFormer is proposed, which models spatio-temporal visual-linguistic dependencies with a static branch and a dynamic branch that performs cross-modal understanding in a single frame and learns to localize the target object spatially according to intra-frame visual cues like object appearances.
- Part Relational Mean Model for Group Re-Identification6
A part-based minus-average relational and arithmetic mean descriptor (PRM) algorithm is proposed to obtain a robust representation of groups to solve the G-ReID problem caused by changes in the number of group members and their relative positions within the group.
- 6
- Adaptive Interaction Modeling via Graph Operations Search6
This paper proposes to search the network structures with differentiable architecture search mechanism, which learns to construct adaptive structures for different videos to facilitate adaptive interaction modeling, and experimentally demonstrates that the designed basic graph operations in the search space are able to model different interactions in videos.
- 6
- 6
- 6
- 6
- Complete Gradient Face: A Novel Illumination Invariant Descriptor6
A novel gradient based descriptor, namely Complete Gradient Face (CGF), is proposed to compensate the limitations in [13] and contribute in three folds to alleviate the illumination effect and enhance facial information based on the Lambertian assumption.
- 6
- 5
- Digitalization of surgical features improves surgical accuracy via surgeon guidance and robotization5
MetaS’s guidance and a lens caliper enabled a surgical robot to perform precise capsulorhexis autonomously in porcine eyes and paves the way for the autonomous operation of surgical robots.
- 5
- ZeroDexGrasp: Zero-Shot Task-Oriented Dexterous Grasp Synthesis with Prompt-Based Multi-Stage Semantic Reasoning5
Experimental results demonstrate that ZeroDexGrasp enables high-quality zero-shot dexterous grasping on diverse unseen object categories and complex task requirements, advancing toward more generalizable and intelligent robotic grasping.
- 5
- 5
- InTraGen: Trajectory-controlled Video Generation for Object Interactions5
This work introduces InTraGen, a pipeline for improved trajectory-based generation of object interaction scenarios, and introduces a multi-modal interaction encoding pipeline with an object ID injection mechanism that enriches object-environment interactions.
- Exploiting Discrepancy in Feature Statistic for Out-of-Distribution Detection5
This paper proposes a simple yet novel approach to OOD detection by leveraging the phenomenon that the average of feature vector elements from convolutional neural network (CNN) is typically larger for ID data than for OOD data.
- 5
- 5
- 5
- Mixed Supervision for Instance Learning in Object Detection with Few-shot Annotation5
This work proposes the Instance-Annotation Guided Image Classification strategy to provide explicit guidance from instance-level annotations by using positional relation to force the image classifier to focus on the proposals which contain the correct object.
- 5
- Guest Editorial Introduction to the Special Issue on Large-Scale Visual Sensor Networks: Architectures and Applications5
In a sensor network, a key role is played by largescale visual networks: the decreasing cost of smart devices enables sensor networks to scale to a large number of devices.
- 5
- 5
- 5
- 5
- Robust large margin discriminant tangent analysis for face recognition5
A robust large margin discriminant tangent analysis method for face recognition that is more robust to outliers and unbalance problems than other nonparametric methods.
- Decoupled Seg Tokens Make Stronger Reasoning Video Segmenter and Grounder4
DeSa2VA is proposed, a decoupling-enhanced prompting scheme integrating text pre-training and a linear decoupling module to address the information processing limitations inherent in SAM-2.
- Revisit the Imbalance Optimization in Multi-task Learning: An Experimental Analysis4
A systematic experimental analysis of the factors contributing to imbalance in multi-task learning suggests that understanding and controlling gradient dynamics is a more direct path to stable MTL than developing increasingly complex methods.
- 4
- 4
- FA: Forced Prompt Learning of Vision-Language Models for Out-of-Distribution Detection4
This study proposes an innovative CLIP-based framework based on Forced prompt leArning (FA), designed to make full use of the InDistribution (ID) knowledge and ultimately boost the effectiveness of OOD detection.
- CoopDiff: Anticipating 3D Human-object Interactions via Contact-consistent Decoupled Diffusion4
This work proposes a novel contact-consistent decoupled diffusion framework CoopDiff, which employs two distinct branches to decouple human and object motion modeling, with the human-object contact points as shared anchors to bridge the motion generation across branches.
- SynopGround: A Large-Scale Dataset for Multi-Paragraph Video Grounding from TV Dramas and Synopses4
A large-scale video grounding dataset named SynopGround is presented, in which more than 2800 hours of videos are sourced from popular TV dramas and are paired with accurately localized human-written synopses, and a novel Local-Global Multimodal Reasoner (LGMR) is proposed to explicitly model the local-global structures of long-term multimodal inputs for MPVG.
- 4
- 4
- 4
- 4
- Devil's in the Detail: Graph-based Key-point Alignment and Embedding for Person Re-ID4
An end-to-end ReID method that integrates both key recognition patterns into a unified framework, called Siamese-GCN, and achieves state-of-the-art performance on three public datasets is proposed.
- Contour-Guided Person Re-identification4
This paper introduces a new data modality, pedestrian contour, into the re-ID community, which to the best knowledge is the first attempt to utilize contour explicitly in deep re-IDs models, and proposes several contour-guided architectures to explicitly use pedestrian contours.
- 4
- 4
- Correlation Based Identity Filter: An Efficient Framework for Person Search4
Extensive experiments illustrate that the CIF framework has the important practical benefit of searching a specific person with a light weight and high efficiency and achieves better accuracy than doing detection and re-identification separately.
- A fast convex conjugated algorithm for sparse recovery4
This paper induces a family of convex conjugated loss functions as a smooth approximation of l(0)-norm to reformulate the sparse recovery problem as an augmented quadratic constraint problem that can be efficiently computed by alternate minimization.
- Unsupervised Selective Transfer Learning for Object Recognition4
A novel unsupervised transfer learning framework that utilises unlabelled auxiliary data to quantify and select the most relevant transferrable knowledge for recognising a target object class from the background given very limited training target samples is proposed.
- 4
- Self-supervised Texture Filtering3
This article presents a self-supervised framework for smoothing out textures while maintaining the image structures and proposes to learn texture filtering from unlabeled data by encouraging the texture inverted image generated from the filtering output to be visually more similar to the input via contrastive learning.
- When Shadow Removal Meets Intrinsic Image Decomposition: A Joint Learning Framework Using Unpaired Data3
By properly coupling shadow removal and IID into a joint learning framework, they can reinforce each other and enable promising results on both tasks, even with unpaired training data.
- 3
- CLIP-RestoreX: Restore Image Structure and Perception in Exposure Correction3
This work presents a novel exposure correction method, referred to as CLIP-RestoreX, that leverages structural and perceptual priors from CLIP to tackle exposure correction, and designs a frequency-domain based feature enhancement diffusion model to better restore the damaged structural information and perceptual information.
- 3
- 3
- Adaptively Integrated Knowledge Distillation and Prediction Uncertainty for Continual Learning3
Two novel ways to adaptively balance model stability and plasticity are proposed, to adaptively integrate multiple levels of old knowledge and transfer it to each block level in the new model and to connect model prediction uncertainty and knowledge distillation for continual learning.
- Learning Discriminative Representation via Metric Learning for Imbalanced Medical Image Classification3
Embedding metric learning into the first stage of the two-stage framework for fine-grained class-imbalanced image classification tasks to help the feature extractor learn to extract more discriminative feature representations consistently outperforms existing onestage and two- stage approaches.
- 3
- 3
- 3
- 3
- 3
- Asymmetric person re-identification: cross-view person tracking in a large camera network3
The development of person RE-ID is reviewed, the recent development on asymmetric distance metric learning and the asymmetric person RE-ID modeling of the largely unsolved open-topic problems is introduced and its challenges are presented.
- A Matrix Splitting Method for Composite Function Minimization3
Extensive experiments show that the proposed and analyzes a new Matrix Splitting Method (MSM) for minimizing composite functions outperforms existing composite function minimization techniques in term of both efficiency and efficacy.
- Learning a Semantically Discriminative Joint Space for Attribute Based Person Re-identification.3
This work proposes to formulate the attribute based person Re-ID as a joint space learning problem, and applies a novel adversary training strategy to generate homogeneous features for both modalities, providing distribution alignment between different modalities in the feature level and keeping semantic consistency across modalities.
- A Similar Interaction Model for Group Activity Recognition in Still Images3
This paper addresses the group activity recognition in the still image with an alternative discriminant contextual model on feature level and shows that the optimization problem can be optimized efficiently.
- Chinese Handwritten Writer Identification based on Structure Features and Extreme Learning Machine.3
Experimental results demonstrate CSF/ELM method can achieve better performance than other traditional schemes for writer identification and learns much faster than traditional popular learning algorithms.
- Learning from Partially Annotated OPT Images by Contextual Relevance Ranking3
A novel learning framework to reduce the requirement of manual annotations while achieving competitive classification performance is presented and the results show that the proposed method can robustly infer patterns from partially annotated images with low computational cost.
- 3
- Structure-Guided Diffusion Models for High-Fidelity Portrait Shadow Removal2
This work first train a shadowindependent structure extraction network on a real-world portrait dataset with various synthetic lighting conditions, which allows to generate a shadow-independent structure map including facial details while excluding the unwanted shadow boundaries.
- 2
- 2
- A Single‐Component Adhesive Sponge Based on Blood‐Triggered and Autopenetrative Adhesion for Robust Vascular Closure2
MonoSeal achieved robust tissue adhesion and effectively sealed the 12 Fr puncture wound on the porcine femoral artery within 30 s, and demonstrates promising performance as a vascular closure device following interventional access in the porcine carotid artery.
- 2
- 2
- Dual-modality adaptation in vision-language models for continual learning2
A novel continual learning framework is proposed that fully leverages the pre-trained vision-language model by simultaneously incorporating task-specific LoRA modules for the image encoder and class-specific learnable text prompts for the text encoder to exploit the potential of the large-scale pre-trained vision-language model.
- Efficient Explicit Joint-level Interaction Modeling with Mamba for Text-guided HOI Generation2
EJIM features a Dual-branch HOI Mamba that separately and efficiently models spatiotemporal HOI information, as well as a Dual-branch Condition Injector for integrating text semantics and object geometry into human and object motions.
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- Learning Multi‐Scale Deep Image Prior for High‐Quality Unsupervised Image Denoising2
This paper presents a novel unsupervised image denoising framework trained from single noisy image without making any explicit assumptions on the noise statistics, built upon the deep image prior (DIP), which enables diverse image restoration tasks.
- 2
- Auxiliary Decoder and Classifier for Imbalanced Skin Disease Diagnosis2
A novel learning framework is proposed to effectively alleviate the impact of class imbalance, adding an auxiliary decoder to reconstruct original images and reusing the original CNN classifier to help the classifier more likely extract disease-relevant features for both rare and common diseases.
- Neighbor Combinatorial Attention for Critical Structure Mining2
This work proposes a novel GCN module named Neighbor Combinatorial ATtention (NCAT) to find critical structure in graph-structured data, which attempts to match combinatorial neighbors with learnable patterns and assigns different weights to each combination based on the matching degree between the patterns and combinations.
- Learning Hierarchical Context for Action Recognition in Still Images2
By propagating information bottom-up in the composition, the model efficiently mines the interaction among local body parts, and integrates those discriminative context cues hierarchically into a compact action representation.
- 2
- One-Shot Learning of Sketch Categories with Co-regularized Sparse Coding2
A transfer learning framework which enables one-shot learning of sketch categories and is based on a novel co-regularized sparse coding model which exploits common/shareable parts among human sketches of seen categories and transfer them to unseen categories.
- Person Re-identification with Data-Driven Features2
A new representation to re-identification that incorporates data-driven features to improve the reliability and robustness in person matching is presented and validated through experiments on two person re-Identification datasets, demonstrating that the proposed representation is effective for person matching.
- Kernel Methods for Facial Image Preprocessing2
This chapter focuses on the pre-image learning problem in KPCA as well as other kernel methods, which is a vital step and also a seriously ill-posed estimation problem for applying kernel methods to image preprocessing.
- Fast Calculation for Fisher Criteria in Small Sample Size Problem2
It will be shown that calculating the eigenvectors of within-class scatter matrix after dimensional reduction can solve the optimal projection for fisher criteria.
- 1
- TacticGen: Grounding Adaptable and Scalable Generation of Football Tactics1
TacticGen is presented, a generative model for adaptable and scalable tactic generation that generates realistic, strategically valuable tactics, demonstrating its practical utility for tactical planning in professional football.
- 1
- Distilling the Unknown to Unveil Certainty1
A flexible framework for OOD knowledge distillation that extracts OOD-sensitive information from a network to develop a binary classifier capable of distinguishing between ID and OOD samples in both scenarios, with and without access to training ID data is presented.
- 1
- 1
- 1
- 1
- Learning Implicit Features with Flow Infused Attention for Realistic Virtual Try-On1
FIA-VTON is proposed, which leverages the implicit warp feature by adopting a Flow Infused Attention module on virtual try-on to enhance implicit warp guidance and incorporate high-level spatial attention to complement the dense warp.
- 1
- Weakly supervised action anticipation without object annotations1
A weakly supervised method is developed that integrates global motion and local finegrained features from current action videos to predict next action label without the need for specific scene context labels and constructs a graph convolutional network for exploiting the inherent relationships of humans and objects under present incidents.
- 1
- Discriminative Distillation to Reduce Class Confusion in Continual Learning1
It is argued that the class confusion phenomena may also play a role in downgrading the classification performance during continual learning, i.e., the high similarity between new classes and any previously learned classes would also cause the classifier to make mistakes in recognizing these old classes, even if the knowledge of theseold classes is not forgotten.
- 1
- 1
- 1
- 1
- On large appearance change in visual tracking1
This paper aims to build a long-term appearance model for robust tracking against large appearance change in two new respects: using historical and distinguishing cues to model target representation and extracting effective spatial objectness features from each frame to distinguish outliers.
- 1
- 1
- 1
- 1
- 1
- 1
- Face Recognition Based on Fusion of High-Frequency and Low-Frequency Components1
The proposed method is applied in the Yale B and acquires the restored illuminated face image with good performance in visual effect and recognition.
- Active Shape Model Based on Sparse Representation1
This paper presents a sparse ASM based on l1-minimization for shape alignment, which can automatically select an effective group of principal components to represent a given shape.
- A Sparse Local Feature Descriptor for Robust Face Recognition1
A new sparse local feature descriptor, namely SLFD, is proposed based on the analysis of SIFT and shown to outperforms the SIFT, LBP based methods and also some other existing face recognition algorithms in terms of recognition accuracy.
- 1
- The Solution Space for Fisher Discriminant Analysis and the Uniqueness Under Constraints1
A theorem is proposed which shows the feasible for constructing the projection, with a simple way to process the construction from the traditional LDA Experiment result shows the ability and feasible of the proposed solution.
- –
- –
- DataClawBench: An Agent Benchmark for Exploratory Real-World Financial Data Analysis–
A systematic evaluation of eight advanced LLMs under the OpenClaw agent reveals that exploratory data analysis breaks agent reliability: more exploration does not reliably translate into task-relevant progress or correct final answers.
- –
- –
- Generative Texture Filtering–
The core idea is to empower texture filtering by taking full advantage of the strong learned image prior of pre-trained generative models via a two-stage strategy to fine-tune a pre-trained generative model via a two-stage strategy.
- –
- –
- –
- Calibration-free Plug-and-Play EEG-based BCIs–
The potential of LOTL to facilitate the deployment of BCI applications under conditions where the source and target domains share a consistent channel configuration is suggested, suggesting the potential of LOTL to facilitate the deployment of BCI applications under conditions where the source and target domains share a consistent channel configuration.
- –
- –
- –
- –
- –
- –
- –
- –
- Revealing Distribution Discrepancy by Sampling Transfer in Unlabeled Data–
A novel approach to address the challenge of test label unavailability is presented, enabling distribution discrepancy evaluation using only training samples, and accurately quantifies the distribution discrepancy, as evidenced by a wide range of complex data scenarios and tasks.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Aggregating Spatio-temporal Context for Video Object Segmentation–
This work exploits the spatio-temporal relationship among image regions by modelling the dependencies among the corresponding visual features with a spatio -temporal RNN to aggregate spatio, temporal, and spatial contextual information for video object segmentation.
- –
- Low-Resolution Person Re-identification by a Discriminative Resolution-Invariant Network–
A novel Discriminative Resolution-invariant Network (DRINet) is proposed to explore the subspace where LR and HR features are highly correlated and the DRINet can extract discriminant features in the commonly shared feature space.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Learning 3D Compact Binary Descriptor for Human Action Recognition in Video–
A learning-based 3D compact binary descriptor (3D-CBD) for human action video representation is presented, a 3D extension of the compact binary face descriptor (CBFD).
- The Methods of Modeling the Image Sets Based on MEAP and Its Application on Face Recognition–
A novel clustering method in the image set-based face recognition, called the Muti-Exemplar Affinity Propagation algorithm (MEAP), which constructs a two-level mapping: ϕ 1 between the feature points and the exemplars, and ϕ 2 between the exemplar and the super-exemplars.
- Main Dialect Identification in Mainland China, Hong Kong and Taiwan–
An improved method based on the analysis of GMM, namely, maximum KL distance based Gaussian component selection (MKLD-GCS) is proposed in order to improve the performance of dialect identification between Mainland China Mandarin and Taiwan Mandarin.
- Asymmetric Facial Shape Based on Symmetry Assumption–
A reflectance model is proposed to simulate the interaction of light and the facial surface under the assumption that face is strictly axial symmetry and works well not only in fitting the intensities of pixel but also in processing the DC component contained in the image.
- –
- –
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.