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
- David SuterSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works303 from public data
- Diverse Embedding Expansion Network and Low-Light Cross-Modality Benchmark for Visible-Infrared Person Re-identification321
A novel augmentation network in the embedding space, called diverse embedding expansion network (DEEN), which can effectively generate diverse embeddings to learn the informative feature representations and reduce the modality discrepancy between the VIS and IR images.
- Towards a Unified Middle Modality Learning for Visible-Infrared Person Re-Identification220
A non-linear middle modality generator (MMG) is proposed, which helps to reduce the modality discrepancy in visible-infrared person re-identification (VI-ReID), and amiddle modality network (MMN) is presented to further enhance the discrimination and richness of features in an explicit manner.
- Feature Decomposition and Reconstruction Learning for Effective Facial Expression Recognition213
Experimental results show that the proposed FDRL method consistently achieves higher recognition accuracy than several state-of-the-art methods, which clearly highlights the benefit of feature decomposition and reconstruction for classifying expressions.
- 189
- A consensus-based method for tracking: Modelling background scenario and foreground appearance179
The proposed method computes SAmple CONsensus (SACON) of the background samples and estimates a statistical model of thebackground, per pixel, which is employed to segment and track people through occlusions.
- Real-Time High-Performance Semantic Image Segmentation of Urban Street Scenes147
This paper proposes a real-time high-performance DCNN-based method for robust semantic segmentation of urban street scenes, which achieves a good trade-off between accuracy and speed.
- 122
- 115
- 111
- 104
- 96
- Incremental Learning of 3D-DCT Compact Representations for Robust Visual Tracking95
This paper constructs an appearance model using the 3D discrete cosine transform (3D-DCT) and designs a discriminative criterion to evaluate the likelihood of a test sample belonging to the foreground object that is embedded into a particle filtering framework for object state inference over time.
- MDPE: A Very Robust Estimator for Model Fitting and Range Image Segmentation95
A novel and highly robust estimator, called MDPE1 (Maximum Density Power Estimator), which applies nonparametric density estimation and density gradient estimation techniques in parametric estimation (“model fitting”).
- 86
- 84
- Linear discriminant analysis using rotational invariant L1 norm78
A novel rotational invariant L"1 norm (i.e., R"1norm) based discriminant criterion (referred to as DCL"1) is proposed, which better characterizes the intra-class compactness and the inter-class separability by using the rotations of the Frobenius norm.
- 78
- 76
- Multi-label learning based deep transfer neural network for facial attribute classification75
A novel deep transfer neural network method based on multi-label learning for facial attribute classification, termed FMTNet, which consists of three sub-networks: the Face detection Network (FNet), the Multi-labelLearning Network (MNet) and the Transfer learning Network (TNet).
- Deep Disturbance-Disentangled Learning for Facial Expression Recognition74
This paper proposes a novel Deep Disturbance-disentangled Learning (DDL) method for FER that is capable of simultaneously and explicitly disentangling multiple disturbing factors by taking advantage of multi-task learning and adversarial transfer learning.
- A fast face detection method via convolutional neural network73
Experimental results on several popular face detection datasets show the efficiency and the effectiveness of the proposed fast face detection method based on discriminative complete features extracted by an elaborately designed convolutional neural network.
- Automatic Image Cropping for Visual Aesthetic Enhancement Using Deep Neural Networks and Cascaded Regression73
A novel cascaded cropping regression (CCR) method to perform image cropping by learning the knowledge from professional photographers is proposed, which improves the convergence speed of the cascaded method, which directly uses random-ferns regressors.
- Stage-Aware Feature Alignment Network for Real-Time Semantic Segmentation of Street Scenes71
A novel Stage-aware Feature Alignment Network (SFANet) based on the encoder-decoder structure for real-time semantic segmentation of street scenes is presented and an auxiliary training strategy is developed to explicitly alleviate the multi-scale object problem without bringing additional computational costs during the inference phase.
- A Novel Robust Statistical Method for Background Initialization and Visual Surveillance70
A novel robust method for the background initialization that can tolerate more than 50% of foreground pixels and noise and achieve very promising results in background initialization: including applications in video segmentation, visual tracking and surveillance.
- Long-Tailed Visual Recognition via Self-Heterogeneous Integration with Knowledge Excavation65
This work proposes a novel MoE-based method called Self-Heterogeneous Integration with Knowledge Excavation (SHIKE), and proposes Depth-wise Knowledge Fusion (DKF) to fuse features between different shallow parts and the deep part in one network for each expert, which makes experts more diverse in terms of representation.
- End-to-End Learning of Object Motion Estimation from Retinal Events for Event-Based Object Tracking65
A novel deep neural network is proposed to learn and regress a parametric object-level motion/transform model for event-based object tracking, which effectively encodes the spatio-temporal information of asynchronous retinal events into TSLTD frames with clear motion patterns.
- Deep Multi-Task Multi-Label CNN for Effective Facial Attribute Classification64
A novel deep multi-task multi-label CNN, termed DMM-CNN, is proposed for effective FAC, which jointly optimizes two closely-related tasks to improve the performance of FAC by taking advantage of multi- task learning.
- Smooth Foreground-Background Segmentation for Video Processing63
This work proposes an efficient way to account for spatial smoothness in foreground-background segmentation of video sequences with a Markov random field, in such a way that it can be globally optimized at video frame rate.
- 62
- FEDIC: Federated Learning on Non-IID and Long-Tailed Data via Calibrated Distillation59
This paper studies the joint problem of non-IID and long-tailed data in federated learning and proposes a corresponding solution called Federated Ensemble Distillation with Imbalance Calibration (FEDIC), a new distillation method with logit adjustment and calibration gating network to solve the long-tail problem effectively.
- 59
- Perspective n-View Multibody Structure-and-Motion Through Model Selection59
This work gives a solution for sequences of several images, full perspective projection, and general rigid motion based on Monte-Carlo sampling and clustering of two-view motions, linking them through the sequence, and model selection to yield the best explanation for the entire sequence.
- Asynchronous Tracking-by-Detection on Adaptive Time Surfaces for Event-based Object Tracking58
An Adaptive Time-Surface with Linear Time Decay (ATSLTD) event-to-frame conversion algorithm, which asynchronously and effectively warps the spatio-temporal information of asynchronous retinal events to a sequence of ATSLTD frames with clear object contours to perform accurate and efficient object tracking.
- Searching for Representative Modes on Hypergraphs for Robust Geometric Model Fitting58
A simple and effective geometric model fitting method to fit and segment multi-structure data even in the presence of severe outliers and achieves significant superiority over several state-of-the-art model fitting methods on both synthetic data and real images.
- 57
- PARFormer: Transformer-Based Multi-Task Network for Pedestrian Attribute Recognition55
The proposed PARFormer network is a pure transformer-based multi-task PAR network consisting of four modules that interact and jointly learn a highly discriminative feature space and supervise the generation of the final features.
- 55
- 55
- 54
- Color Image Segmentation Using Global Information And Local Homogeneity52
This paper proposes a new method of color image segmentation considering both global information and local homogeneity in HSV, which applies the mean shift algorithm in the hue and intensity subspace of HSV.
- Multi-subregion based correlation filter bank for robust face recognition51
The proposed Multi-Subregion based Correlation Filter Bank (MS-CFB) combines the benefits of global-based and local-based feature extraction algorithms, where multiple correlation filters corresponding to different face subregions are jointly designed to optimize the overall correlation outputs.
- Joint Deep Learning of Facial Expression Synthesis and Recognition50
A novel joint deep learning of facial expression synthesis and recognition method for effective FER and an intra- class loss with a novel real data-guided back-propagation (RDBP) algorithm to reduce the intra-class variations of images from the same class, which can significantly improve the final performance.
- 50
- Effective Appearance Model and Similarity Measure for Particle Filtering and Visual Tracking50
Experiments show that the adaptively model the appearance of objects based on Mixture of Gaussians in a joint spatial-color space can successfully track objects despite changes in foreground appearance, clutter, occlusion, etc.
- 48
- Low-resolution facial expression recognition: A filter learning perspective47
Experimental results on several facial expression datasets show the superior performance of the proposed IFSL method for low-resolution facial expression recognition, compared with several state-of-the-art methods.
- A Model-Selection Framework for Multibody Structure-and-Motion of Image Sequences47
A framework for multibody structure-and-motion based on model selection is presented and can be used with any parametric camera model, or with a combination of different models, and it is robust to realistic amounts of outliers.
- Adaptive Deep Disturbance-Disentangled Learning for Facial Expression Recognition45
Extensive experimental results show that the proposed ADDL method consistently outperforms several state-of-the-art FER methods, demonstrating the great potential of disturbance disentanglement for FER.
- Semantic-Guided Relation Propagation Network for Few-shot Action Recognition44
A novel semantic-guided relation propagation network (SRPN), which leverages semantic information together with visual information for few-shot action recognition and outperforms several state-of-the-art methods with a significant margin.
- The Ordered Residual Kernel for Robust Motion Subspace Clustering44
A novel and highly effective approach for multi-body motion segmentation which encapsulate the hypotheses in a novel Mercer kernel which elicits the potential of two point trajectories to have emerged from the same subspace.
- Single image super-resolution with multi-scale information cross-fusion network43
A multi-scale information cross-fusion network (MSICF) in which a sequence of subnetworks is cascaded to infer high resolution features in a coarse-to-fine manner is proposed to improve information flow and capture sufficient knowledge for reconstructing high-frequency details.
- Label-Noise Learning with Intrinsically Long-Tailed Data42
This paper proposes two-stage bi-dimensional sample selection (TABASCO), a learning frame-work for label-noise learning with intrinsically long-tailed data to better separate clean samples from noisy samples, especially for the tail classes.
- Deep Multi-Branch Aggregation Network for Real-Time Semantic Segmentation in Street Scenes42
A novel Deep Multi-branch Aggregation Network based on the encoder-decoder structure to perform real-time semantic segmentation in street scenes is proposed and extensive experimental results demonstrate that DMA-Net provides a good tradeoff between segmentation quality and speed.
- 42
- 41
- Multi-task Learning of Cascaded CNN for Facial Attribute Classification40
A novel multi-task learning of cascaded convolutional neural network method, termed MCFA, for predicting multiple facial attributes simultaneously, which outperforms several state-of-the-art FAC methods on the challenging CelebA and LFWA datasets.
- 38
- 37
- Using symmetry in robust model fitting37
This paper adapts the concept of "symmetry distance" to formulate an improved regression method, called the least trimmed symmetry distance (LTSD), and shows that the LTSD performs better than LMedS and LTS under a large percentage of clustered outliers and large standard variance of inliers.
- 35
- Dual Semantic Fusion Network for Video Object Detection35
This work proposes a dual semantic fusion network (abbreviated as DSFNet) to fully exploit both frame-level and instance-level semantics in a unified fusion framework without external guidance and introduces a geometric similarity measure into the fusion process to alleviate the influence of information distortion caused by noise.
- 34
- 34
- Toward Video-Based Navigation for Endoscopic Endonasal Skull Base Surgery34
This work proposes a method to directly track the position of the endoscope using video data, which presents registration results within 1mm, which matches the accuracy of the validation framework.
- 33
- Hypergraph modelling for geometric model fitting33
A novel hypergraph based method to fit and segment multi-structural data and includes a robust hypergraph partition algorithm that is superior to some state-of-the-art fitting methods.
- Mode-Seeking on Hypergraphs for Robust Geometric Model Fitting33
A novel geometric model fitting method, called Mode-Seeking on Hypergraphs (MSH), to deal with multi-structure data even in the presence of severe outliers and has significant superiority over the state-of-the-art fitting methods on both synthetic data and real images.
- When Facial Expression Recognition Meets Few-Shot Learning: A Joint and Alternate Learning Framework29
This paper proposes a novel Emotion Guided Similarity Network (EGS-Net), consisting of an emotion branch and a similarity branch, based on a two-stage learning framework, which aims to identify unseen compound expressions with the model trained on easily accessible basic expression datasets.
- Robust visual tracking via spatio-temporal adaptive and channel selective correlation filters29
Spatio-Temporal adaptive and Channel selective Correlation Filters (STCCF) can not only alleviate the over-fitting problem and reduce the computational cost, but also enhance the discriminability and interpretability of the learned filters.
- 29
- 28
- 28
- Hypergraph Optimization for Multi-Structural Geometric Model Fitting28
A novel hypergraph optimization based model fitting (HOMF) method to construct a simple but effective hypergraph that outperforms several state-of-the-art model fitting methods on both synthetic data and real images, especially in sampling efficiency and in handling data with severe outliers.
- 26
- Robust Fitting by Adaptive-Scale Residual Consensus26
A new estimator called Adaptive-Scale Residual Consensus (ASRC), which can simultaneously estimate the parameters of a model and the scale of inliers belonging to that model and has better robustness to heavily corrupted data than other robust methods.
- 24
- 24
- 23
- 23
- 23
- 23
- 22
- 22
- Revisiting graph construction for fast image segmentation22
This paper re-examine the locality-preserving character of spectral clustering by constructing a graph over image regions with both global and local connections, and shows that the generated eigenvector histogram based representation can automatically drive effective unary potentials for a hierarchical random field model to produce multi-class segmentation.
- Multiscale Cascaded Scene-Specific Convolutional Neural Networks for Background Subtraction22
This paper proposes a multiscale cascaded scene-specific CNNs based background subtraction method equipped with a novel training strategy, which takes advantage of the balance of positive and negative training samples to effectively train the robust neural network models.
- 21
- 21
- Adaptive Middle Modality Alignment Learning for Visible-Infrared Person Re-identification20
An Adaptive Middle-modality Alignment Learning (AMML) method, which can effectively reduce the modality discrepancy via an adaptive middle modality learning strategy at both image level and feature level, is proposed.
- 20
- 19
- 19
- Segmentation by Continuous Latent Semantic Analysis for Multi-structure Model Fitting19
A new latent semantic space is constructed, where inliers of different model instances are mapped into several independent directions, while gross outliers are distributed close to the origin of LSS, to efficiently and effectively estimate the parameters of model instances in data.
- 19
- Personalized Federated Learning on Long-Tailed Data via Adversarial Feature Augmentation18
FedAFA optimizes the personalized model for each client by producing a balanced feature set to enhance the local minority classes and significantly improves the personalized performance of each client compared with the state-of-the-art PFL algorithm.
- 18
- 18
- 18
- 18
- Superpixel-Guided Two-View Deterministic Geometric Model Fitting18
Experimental results show that the proposed SDF has significant advantages over several state-of-the-art fitting methods when it is applied to real images with single-st structure and multiple-structure data.
- False-peaks-avoiding mean shift method for unsupervised peak-valley sliding image segmentation18
This paper shows both empirically and analytically that when using sample data, the reconstructed PDF may have false peaks and how the occurrence of the false peaks is related to the bandwidth h of the kernel density estimator, using examples of gray-level image segmentation.
- 17
- D³Net: Dual-Branch Disturbance Disentangling Network for Facial Expression Recognition17
This paper proposes a novel Dual-branch Disturbance Disentangling Network (D3Net), mainly consisting of an expression branch and a disturbance branch, to perform effective FER and introduces an Indian buffet process (IBP) prior to model the distribution of potential disturbing factors in LFS, thereby enhancing the disentanglement of disturbing factors.
- Robust Heterogeneous Model Fitting for Multi-source Image Correspondences16
A heterogeneous model fitting method is proposed, which can effectively estimate the parameters of the transformation model for alleviating the influence of outliers and outperforms several state-of-the-art competing methods in terms of matching performance.
- 16
- Visual saliency detection based on homology similarity and an experimental evaluation16
A graph-based superpixel-wise similarity called "homology similarity" is proposed, which describes how likely two superpixels belong to the same object or background region and results show that, compared with these competing models, the model yields promising results.
- A model-based range image segmentation algorithm using a novel robust estimator15
A novel range image segmentation algorithm based on a newly proposed robust estimator: Adaptive Scale Sample Consensus (ASSC) is presented, which is a model-based top-down technique and directly extracts the required primitives (models) from the raw images.
- DSNet: Deep and Shallow Feature Learning for Efficient Visual Tracking14
A deep and shallow feature learning network is presented to learn the multi-level same-resolution compressed (MSC) features for efficient online tracking, in an end-to-end offline manner and shows that the learned MSC features have the appealing advantage of allowing the equipped DCF-based tracking methods to perform favorably against the state-of-the-art methods while running at high frame rates.
- Scene text recognition using sparse coding based features14
Experiments demonstrate that the HSC-based scene text recognition method outperforms the HOG-based method significantly and achieves the state-of-the-art performance.
- Robust visual tracking by metric learning with weighted histogram representations14
A robust visual tracking algorithm is proposed that incorporates online distance metric learning into visual tracking based on a particle filter framework and uses spatially weighted feature representations using both color and spatial information of objects, which can further improve the tracking performance.
- Background subtraction driven seeds selection for moving objects segmentation and matting14
This paper proposes a new automatic way to integrate a background subtraction (BGS) and an alpha matting technique via a heuristic seeds selection scheme, and demonstrates the efficiency and effectiveness of this method.
- 13
- 13
- Efficient Semidefinite Spectral Clustering via Lagrange Duality13
Compared with the state-of-the-art spectral clustering methods, the proposed algorithm achieves better clustering performance and is much more efficient and can solve larger-scale SSC problems than those standard interior-point SDP solvers.
- Evaluation of local features and classifiers in BOW model for image classification13
An EMD spatial kernel is proposed to encode the spatial information of local features in the framework of the BOW model and the experimental results show that the proposed kernel outperforms the EMD kernel which does not consider the spatial Information in image classification.
- Human Action Recognition Using Pyramid Vocabulary Tree13
This pape proposes a pyramid vocabulary tree to model local spatio-temporal features, which can characterize the inter-class difference and also allow intra-class variance and results show that both the Pyramid vocabulary tree and the SST-PMK lead to a significant improvement in human action recognition.
- 13
- 12
- 12
- 12
- 12
- Robust Correlation Filter Tracking with Shepherded Instance-Aware Proposals12
A novel CNN based correlation filter tracker with shepherded instance-aware proposals, namely DeepCFIAP, which automatically estimates the target scale in each frame and re-detects the target when distortion happens and develops a novel proposal ranking algorithm based on the similarities between proposals and instances.
- 12
- 12
- 12
- Expression-targeted feature learning for effective facial expression recognition11
A joint loss is developed to explicitly reduce intra-class variations while enlarging inter-class differences to enhance the discriminability of ESF and introduces the sample-sensitive weights and the soft-expression weights to balance the joint loss for better performance.
- 11
- Scene character recognition using PCANet11
The proposed method achieves promising performance on the Chars74K-15 dataset and the ICDAR03-CH dataset, demonstrating the effectiveness of PCANet in scene character recognition.
- Effective Facial Expression Recognition via the Boosted Convolutional Neural Network11
A novel and effective FER via the Boosted Convolutional Neural Network (Boosted-CNN), which can achieve better classification rates compared with other state-of-art methods.
- 11
- Consistency-driven feature scoring and regularization network for visible–infrared person re-identification10
A novel consistency-driven feature scoring and regularization network (CFSR-Net) is proposed, which consists of a backbone network, a local feature learning block, a feature scoring block, and a global-local feature fusion block, for VI-ReID.
- 10
- 10
- 10
- Object Discovery via Cohesion Measurement10
The affinity matrices originally used in image segmentation methods based on spectral clustering are reinvestigate and a novel object discovery method is proposed to discover objects latent in an image by utilizing the eigenvectors of the affinity matrix.
- Robust statistics for computer vision : model fitting, image segmentation and visual motion analysis10
- Anatomical Reconstruction from Endoscopic Images: Toward Quantitative Endoscopy10
This study indicates that processing endoscopic images to produce anatomic structure is feasible and has high potential clinical value for intraoperative navigation, diagnosis, and treatment planning.
- Sacon: A consensus based model for background subtraction10
This paper presents an effective and adaptive background modeling method for detecting foreground objects in both static and dynamic scenes and computes SAmple CONsensus (SACON) of the background samples and estimates a statistical model per pixel.
- Uncertainty-Aware Label Refinement on Hypergraphs for Personalized Federated Facial Expression Recognition9
Experimental results on two challenging real-world facial expression databases show that the proposed method consistently outperforms several state-of-the-art methods, indicating the superiority of hypergraph modeling for uncertainty estimation and label refinement on the personalized federated FER task.
- TRL: Transformer based refinement learning for hybrid-supervised semantic segmentation9
A novel learning method, Transformer based Refinement Learning (TRL), which explores a learning process under the assistance of weak annotations and the supervision of strong annotations for hybrid-supervised semantic segmentation.
- 9
- 9
- Superpixel-Based Two-View Deterministic Fitting for Multiple-Structure Data9
Experimental results demonstrate that the proposed Superpixel-based Deterministic Fitting (SDF) shows superiority over several state-of-the-art fitting methods for real images with single-st structure and multiple-structure data.
- 9
- 9
- Learning Group Activity in Soccer Videos from Local Motion9
A local motion-based approach for recognizing group activities in soccer videos that focuses on local motion of individuals and excludes noise such as background motion caused by inaccurate compensation.
- 8
- You are Your Own Best Teacher: Achieving Centralized-level Performance in Federated Learning under Heterogeneous and Long-Tailed Data8
This paper proposes FedYoYo (You Are Your Own Best Teacher), introducing Augmented Self-bootstrap Distillation (ASD) to improve representation learning by distilling knowledge between weakly and strongly augmented local samples, without needing extra datasets or models.
- 8
- 8
- Semi-supervised Visible-Infrared Person Re-identification via Modality Unification and Confidence Guidance8
The proposed MUCG has significant advantages in improving the performance of the SSVI-ReID task, surpassing the current state-of-the-art methods by a significant margin.
- IMC-Det: Intra–Inter Modality Contrastive Learning for Video Object Detection8
A novel intra–inter modality contrastive learning network for high-performance video object detection (IMC-Det), which includes three substantial improvements over existing methods, and develops a graph relational feature aggregation module to effectively model the structural relations between features by leveraging cross-graph learning and residual graph convolution.
- Augmentation Matters: A Mix-Paste Method for X-Ray Prohibited Item Detection Under Noisy Annotations8
This paper addresses the challenging problem of training a robust prohibited item detector under noisy annotations from a novel perspective of data augmentation, and proposes an effective label-aware mixed patch paste augmentation method (Mix-Paste), which demonstrates the great potential of data augmentation to handle noise annotations.
- SPL-Net: Spatial-Semantic Patch Learning Network for Facial Attribute Recognition with Limited Labeled Data8
Experimental results on challenging facial attribute datasets show the superiority of SPL-Net over several state-of-the-art methods in the case of limited labeled data.
- 8
- Optical flow estimation via weighted guided filtering with non-local steering kernel8
The proposed weighted guided filtering with a non-local steering kernel is robust for optical flow estimation and able to preserve motion boundaries and is compared with several state-of-the-art methods using the Middlebury and MPI Sintel test datasets.
- 8
- 8
- Mode seeking on graphs for geometric model fitting via preference analysis8
It is shown that the proposed graph-based mode-seeking method has significant superiority for geometric model fitting, which enables mode seeking to be beneficial for reducing the influence of bad hypotheses since bad hypotheses usually have larger residual values than good ones.
- A unified hypothesis generation framework for multi-structure model fitting8
A Unified Hypothesis Generation (UHG) framework is presented, which effectively combines the conventional multi-structure hypothesis generation strategy with the proposed guided sampling strategy by using a Markov Chain Monte Carlo process based on a cooling schedule.
- Efficient guided hypothesis generation for multi-structure epipolar geometry estimation8
Experimental results on public real image pairs show that EGHG significantly outperforms several state-of-the-art sampling methods on multi-structure data.
- Quadratic projection based feature extraction with its application to biometric recognition8
Experimental results demonstrate the superiority of the DualQML-based feature extraction algorithm compared to the current state-of-the-art algorithms on biometric recognition tasks.
- Recent Advances on Supervised Distance Metric Learning Algorithms8
Recently developed algorithms for supervised distance metric learning based on the partition of pairwise constraints and non-pairwise constraints are reviewed and some representative algorithms are introduced and their respective pros and cons are analyzed.
- Robust Object Tracking in Crowd Dynamic Scenes Using Explicit Stereo Depth8
The proposed depth-driven tracking approach can largely alleviate the drifting issue, especially when the object frequently interacts with similar background in long sequence tracking, and the problems caused by rapid object appearance change can be avoided due to the stability of the depth cue.
- An effective unconstrained correlation filter and its kernelization for face recognition8
A non- linear extension of UOOTF based on the kernel technique allows for higher flexibility of the decision boundary due to a wider range of non-linearity properties and to handle non-linearly separable distributions between different classes.
- Boosting histograms of descriptor distances for scalable multiclass specific scene recognition8
A scheme based on Multiclass AdaBoost to select the appropriate reference keypoints and the result is a scalable multiclass specific scene classifier capable of processing a large number of scene classes at a fraction of the time required for methods based on exhaustive keypoint matching.
- 7
- 7
- 7
- Robust optical flow estimation to enhance behavioral research on ants7
An edge-preserving filter-based optical flow method is constructed to estimate the motion of ants in outdoor complex scenarios and the results reveal that the proposed method can improve the accuracy of motion estimation.
- Learning Object Scale With Click Supervision for Object Detection7
This work proposes a simple yet effective method which incorporates CNN visualization with click supervision to generate pseudo ground-truths, which can be used to train a fully-supervised detector.
- Affine hull based target representation for visual tracking7
This work proposes a novel target representation for tracking that considers both the importance of one template to a target candidate in the current frame and the recentness of the template that is kept in the template set.
- 6
- 6
- 6
- 6
- 6
- 6
- Scene Character and Text Recognition: The State-of-the-Art6
The state-of-the-art performance is provided to show the achieved performance so far and demonstrate the potential of deep learning based methods.
- Learning Hough regression models via bridge partial least squares for object detection6
A novel Hough Transform-based object detection approach to establish context-encoded Hough Regression Models (HRMs), which are linear regression models that cast probabilistic Hough votes to predict object locations and a novel multi-scale voting scheme to efficiently handle object scale changes.
- Robust object tracking using a spatial pyramid heat kernel structural information representation6
This paper takes advantage of heat kernel structural information (HKSI) matrices to represent object appearance, because HKSI matrices perform well in characterizing the edge flow information on the object appearance graph.
- 6
- 5
- 5
- 5
- 5
- 5
- 5
- 5
- Large margin deep embedding for aesthetic image classification5
An LMDE method with a novel network structure and an effective joint loss function, which takes advantage of both the triplet loss function and the hinge loss function is presented.
- Fast Nearest Neighbor Search Based on Approximate k-NN Graph5
A novel k-nearest neighbor graph construction method that makes the best trade-off between search quality and computational cost is proposed.
- 5
- Rapid hypothesis generation by combining residual sorting with local constraints5
Experimental results on homography and fundamental matrix estimation show that RHG can effectively guide hypothesis generation process and rapidly generate promising hypotheses for heavily contaminated multi-structure data.
- A Solution to Co-occurrence Bias in Pedestrian Attribute Recognition: Theory, Algorithms, and Improvements4
A novel concept of attributes-disentangled feature learning is formulated, by which the mutual information among features of different attributes is minimized, ensuring the recognition of an attribute independent to the presence of others.
- Classifying Long-tailed and Label-noise Data via Disentangling and Unlearning4
This work investigates the dynamic misleading process of the nosiy labels and proposes a novel method called Disentangling and Unlearning for Long-tailed and Label-noisy data (DULL), which prevents the model from being misled by noisy labels, enhancing the model's robustness against noise.
- Diverse Consensuses Paired with Motion Estimation-Based Multi-Model Fitting4
A novel method of diverse Consensuses paired with Motion estimation-based multi-Model Fitting (CMMF), which leverages three types of diverse consensuses along with inter-model collaboration to enhance the effectiveness of multi-model fusion.
- 4
- Explicit State Representation Guided Video-based Pedestrian Attribute Recognition4
This paper proposes the Explicit State Representation (ExSR) Guided Pedestrian Attribute Recognition to improve the accuracy through state learning and attribute fusion among frames and demonstrates that the ExSR outperforms the state-of-the-art methods on two public databases.
- 4
- Learning intra-inter semantic aggregation for video object detection4
In this paper, a novel intra-inter semantic aggregation method (ISA) is proposed to learn more effective intra and inter relations for semantically aggregating object features and can generate discriminative features from the novel perspective of intra- inter semantic aggregation for robust video object detection.
- 4
- 4
- 4
- A Novel Unconstrained Correlation Filter and Its Application in Face Recognition4
A novel unconstrained correlation filter called Unconstrained Optimal Origin Tradeoff Filter (UOOTF) is presented and applied to face recognition, which increases the overall performance for unseen patterns by removing the hard constraints on the outputs during the filter design.
- Class-Specified Segmentation with Multi-scale Superpixels4
A class-specified segmentation method, which can not only segment foreground objects from background at pixel level, but also parse images, which not only has fewer free parameters but also is simpler and effective.
- Occlusion Handling with ℓ1-Regularized Sparse Reconstruction4
This paper proposes a novel framework to track multi-object with occlusion handling according to sparse reconstruction that can automatically focus on the visible part of the occluded object, and thus exclude the need of deducing the Occlusion relationship.
- 3
- FATE: A Prompt-Tuning-Based Semi-Supervised Learning Framework for Extremely Limited Labeled Data3
FATE effectively mitigates challenges arising from the scarcity of labeled samples in SSL, achieving an average performance improvement of 33.74% across seven benchmarks compared to state-of-the-art SSL methods.
- Federated Semi-Supervised Learning with Annotation Heterogeneity3
This paper proposes a novel FSSL framework called heterogeneously annotated semi-supervised learning (HASSLE), which employs a dual-model approach and proves that the dual models in HASSLE learned by SUMA can mutually learn from each other, thereby effectively utilizing the information of both types of data across different clients.
- 3
- 3
- 3
- 3
- DPTNet: A Dual-Path Transformer Architecture for Scene Text Detection3
This paper proposes DPTNet (Dual-Path Transformer Network), a simple yet effective architecture to model the global and local information for the scene text detection task, and achieves state-of-the-art results on the MSRA-TD500 dataset, and provides competitive results on other standard benchmarks in terms of both detection accuracy and speed.
- Joint regularization and low-rank fusion for atmospheric turbulence removal3
Extensive experimental results show that the novel strategy presented can effectively mitigate geometric deformation as well as blur variations and that it outperforms several other state-of-the-art turbulence removal methods.
- Small-Vote Sample Selection for Label-Noise Learning3
This paper proposes a novel yet simple sample selection method, which mainly consists of a Hierarchical Voting Scheme (HVS) and an Adaptive Clean data rate Estimation Strategy (ACES), to accurately identify clean samples and noisy-labeled samples for robust learning.
- A two-step hypergraph reduction based fitting method for unbalanced data3
A two-step hypergraph reduction based fitting for unbalanced data by extending the binary incident matrix of a normal hypergraph to a continuous (soft) generalization, to improve the accuracy of hypergraph partition for model fitting.
- A Novel Background Subtraction Method Based on ViBe3
The proposed method can effectively alleviate the problems caused by foreground aperture, dynamic backgrounds and ghosts in background subtraction, and achieves competitive results against several state-of-the-art methods.
- 3
- Robust geometric model fitting based on iterative Hypergraph Construction and Partition3
Experimental results on real images show the superiority of the proposed IHCP method over several state-of-the-art model fitting methods.
- Message Passing on the Two-Layer Network for Geometric Model Fitting3
A novel model fitting method to recover multiple geometric structures from data corrupted by noises and outliers is proposed, which combines both the consensus information in all model hypotheses and the preference Information in all data points into a two-layer network.
- A Survey of Background Modeling Based on Robust Subspace Learning via Sparse and Low-rank Matrix Decomposition3
This paper surveys the background modeling methods based on robust subspace learning via sparse and low-rank matrix decomposition in terms of different sub space learning formulation frameworks, and then analyzes their respective pros and cons.
- 3
- 3
- 3
- Video-to-Task Learning via Motion-Guided Attention for Few-Shot Action Recognition2
A novel Dual Motion-Guided Attention Learning method (called DMGAL) is proposed, aiming to capture the spatio-temporal relationships from the video-specific level to the task-specific level, to achieve spatio-temporal relationships at the task level.
- Adaptive Graph Attention-Guided Parallel Sampling and Embedded Selection for Multi-Model Fitting2
A dual-confidence graph attention module that models data relationships using an adaptive graph attention network and a quality-embedded selection module that integrates quality confidence into the joint optimization of model selection and data clustering are presented.
- Wavelet-domain feature decoupling for weakly supervised multi-object tracking2
Experimental results show the superiority of FDMOT compared with several state-of-the-art supervised and weakly supervised MOT methods.
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- A Hierarchical Voting Scheme for Robust Geometric Model Fitting2
The proposed Hierarchical Voting scheme based Fitting method can effectively estimate model instances from the model hypotheses generated by random sampling, which usually includes a large proportion of “bad” model hypotheses.
- The State-of-the-Art Research Progress on Motion Segmentation2
This paper reviews the recent developments (especially for those in recent five years) of motion segmentation methods and analyze their respective pros and cons.
- 2
- 2
- Low Rank Metric Learning via Accelerated Gradient Optimization2
To effectively solve the proposed low rank metric learning problem, an efficient algorithm is developed based on the accelerated gradient method that not only learns an effective metric, but also projects data into an intrinsic low-dimensional subspace.
- Robust Head Pose Estimation with a New Principal Optimal Tradeoff Filter2
Based on the correlation filter technique, a novel feature extraction framework, i.e., directional correlation filter set (DCFS), is developed for robust head pose estimation, designed in the feature subspace obtained according to principal component analysis (PCA).
- 2
- 2
- 1
- Joint Implicit and Explicit Language Learning for Pedestrian Attribute Recognition1
A Joint Implicit and Explicit Language Guidance Enhancement Learning (JGEL) method, which converts each pedestrian image into a language description with dual language learning to effectively learn enhanced attribute information, is proposed.
- 1
- 1
- OV-VOD: Open-Vocabulary Video Object Detection1
Experiments on multiple video datasets demonstrate that the proposed OV-VOD exhibits superior zero-shot generalization capability compared to existing image-level open-vocabulary object detection methods.
- TFPA: Text Features Guided Dynamic Parameter Adjustment for Few Shot Action Recognition1
A Text Feature guided dynamic Parameter Adjustment (TFPA) method for few-shot action recognition inspired by basis decomposition in vector spaces, which promotes functional specialization and reduce redundancy across the basis mapping matrices, thereby enhancing performance.
- 1
- 1
- Towards Unseen Triples: Effective Text-Image-joint Learning for Scene Graph Generation1
A Text-Image-joint Scene Graph Generation (TISGG) model to resolve the unseen triples and improve the generalisation capability of the SGG models, and a novel balanced learning strategy, including a Charater Guided Sampling and an Informative Re-weighting module, to provide tailor-made learning methods for each predicate according to their characters.
- 1
- 1
- An End-to-End Scene Text Detector with Dynamic Attention1
A dynamic end-to-end framework which includes a convolution-based dynamic encoder with various attention types to generate a deformable and dynamic view for multi-oriented text instances and curve ones and achieves superior performance compared with several state-of-the-art scene text detectors.
- Global and local feature alignment for video object detection1
A Global and Local Feature Alignment (abbreviated as GLFA) module for video object detection is proposed, which can distill both global and local information to excavate the deep relationship between features for feature alignment.
- Conceptual space based model fitting for multi-structure data1
Conceptual Space based Model Fitting is able to efficiently and effectively remove gross outliers in data, and simultaneously estimate the number and the parameters of model instances without using prior information.
- Adaptive Metric Learning with the Low Rank Constraint1
An Adaptive Metric Learning with the Low Rank Constraint (AML-LR) method, which restricts the learned distances between examples of pairs using adaptive bounds and meanwhile the rank of the learned matrix is minimized, which is adaptive to different data distributions and robust to avoid overfitting is proposed.
- Hierarchical Learning for Large-Scale Image Classification via CNN and Maximum Confidence Path1
This framework proposes a framework to integrate the large scale image data visualization with image classification that achieves the highest top-1 and top-5 classification accuracy in comparison with 6 state-of-the-art methods.
- An Outlier Removal Method by Statistically Analyzing Hypotheses for Geometric Model Fitting1
An outlier removal method which utilizes the information of hypotheses for model fitting is proposed and it is shown that the bad hypotheses, whose parameters are far from the parameters of model instances in data, also contain the correlation information between data points.
- 1
- Text Localization with Hierarchical Multiple Feature Learning1
A kind of two level character structure features in collaboration with the Histogram of Gradient and the Convolutional Neural Network features for character localization and an optimal splitting strategy based on the interval cues to split strings into words are introduced.
- Comparative Study: Face Recognition via the Correlation Filter Technique1
A comparative study by using the correlation filter (CF) technique in the application of face recognition to provide widely applicable guidelines for designing the face recognition systems via the CF technique.
- Visual Saliency Detection via Homology Distribution and Color Contrast1
4th International Conference on Intelligence Science and Big Data Engineering, IScIDE 2013, Beijing, China, July 31 to August 2, 2013.
- 1
- 1
- Non-linear Feature Fusion Based on Polynomial Correlation Filter for Face Recognition1
This study proposes a novel and effective fusion method, which uses polynomial correlation filter (PCF) to non-linearly fuse different types of features for robust face recognition.
- 1
- 1
- –
- Is Spurious Correlation Removal Always Learnable?–
This work shows a conditional computational barrier: under a black-box samplable supervised sparse recovery primitive motivated by average-case sparse-recovery reductions, there exist multi-environment instances with a one-dimensional predictive invariant subspace that are learnable with polynomial samples by exhaustive search, while any polynomial-time constant-accuracy recovery algorithm would contradict the primitive.
- Change Only What You Need: Dual Dynamic Hypergraph-Driven Fusion for GAN Inversion–
This work proposes a novel dual dynamic hypergraph-driven fusion (DHFusion) method, which consists of a dual dynamic hypergraph-CAM network (DH-Net) and an editing-driven fusion network (EF-Net), and shows the potential of dual dynamic hypergraphs for GAN inversion.
- AUCH-Net: Action Unit-Based Consistency-Aware Hypergraph Network for Cross-Domain Few-Shot Facial Expression Recognition–
A novel Action Unit-based Consistency-aware Hypergraph Network (AUCH-Net), which constructs consistency-aware hypergraphs on AUs, for CF-FER and can bridge the gap between fine-grained facial variations and high-level expression categories, greatly facilitating the learning of transferable feature representations.
- –
- –
- –
- VCD-Net: Visual-Textual Cascaded Decomposition Network for Cross-Domain Few-Shot Facial Expression Recognition–
This paper proposes a novel visual-textual cascaded decomposition network (VCD-Net) for compound FER under the cross-domain few-shot learning setting, and demonstrates the superiority of VCD-Net over several state-of-the-art few-shot learning methods.
- –
- –
- –
- –
- –
- –
- DeCAB: Debiased Semi-supervised Learning for Imbalanced Open-Set Data–
DeCAB introduces positive-pair scores for contrastive learning instead of positive-negative pairs based on unreliable pseudo-labels, avoiding false negative pairs negatively impacts the feature space, and utilizes class-aware thresholds to select more tail samples and selective sample reweighting for feature learning.
- –
- –
- –
- –
- –
- –
- –
- –
- Robust visual tracking via scale-aware localization and peak response strength–
A robust online tracking method via Scale-Aware localization and Peak Response strength (SAPR) which can learn a discriminative model predictor to estimate a target state accurately is proposed, achieving the favorable performance against several state-of-the-art trackers.
- –
- Learning Target-Specific Response Attention for Siamese Network Based Visual Tracking–
A Target-Specific Response Attention (TSRA) module is proposed to enhance the discriminability of the Siamese network based visual tracking methods and introduces a shortcut connection strategy to generate a residual multi-channel response map for more discriminative tracking.
- Adaptive Correlation Filter Tracking with Weighted Foreground Representation–
A weighted foreground appearance feature descriptor which effectively characterizes the appearance of objects is presented and an adaptive model updating strategy to mitigate the problem that the models are updated by using corrupted samples is proposed.
- Evolution-based outlier removal for geometric model fitting–
A fitness function is developed to improve the “quality” of selected point sets, which is then used to remove outliers and statistically analyze the evolutional results and distinguish inliers from outliers.
- Induced subgraph searching for geometric model fitting–
The proposed model fitting method based on graphs to fit and segment multiple-structure data is able to effectively estimate the number and the parameters of model instances in data severely corrupted by outliers and noises.
- A Comparative Study On Features for Similar Image Search–
A hybrid feature representation that combines keypoint detector and CNN descriptor is proposed, in which the geometric invariances of keypoint feature and the distinctiveness of CNN feature are integrated and shows that this hybrid feature achieves superior performance.
- STV2k–
A large-scale and well-annotated scene text dataset, namely STV2k, is presented, which can be used for scene text detection as well as scene text recognition.
- –
- Stereo Context for Robust Object Tracking–
A stereo context-based object tracker that explores 3D context information for collaborative tracking in binocular video streams and demonstrates the robustness and superior performance of the proposed tracker.
- Weakly-supervised object localization in unlabeled image collection–
This paper proposes a weakly supervised approach to localize objects in a class of unlabelled images without using any manually labeled examples, and demonstrates that the approach is effective and accurate to automatically label images.
- Spatio-temporal proximity distribution kernels for action recognition–
A spatio-temporal proximity distribution kernel (ST-PDK) is proposed to measure the similarity of videos, which satisfies Mercer condition and is directly incorporated into the kernel function of the SVM classifier.
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.