Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks291 from public data
- Go-ICP: A Globally Optimal Solution to 3D ICP Point-Set Registration1,170
This paper presents the first globally optimal algorithm, named Go-ICP, for Euclidean (rigid) registration of two 3D point-sets under the inline-formula notation, and derives novel upper and lower bounds for the registration error function.
- Accurate 3D Face Reconstruction With Weakly-Supervised Learning: From Single Image to Image Set910
A novel deep 3D face reconstruction approach that leverages a robust, hybrid loss function for weakly-supervised learning which takes into account both low-level and perception-level information for supervision, and performs multi-image face reconstruction by exploiting complementary information from different images for shape aggregation is proposed.
- 517
- 385
- 156
- 155
- Learning Human Interaction by Interactive Phrases140
A novel hierarchical model to encode interactive phrases based on the latent SVM framework where interactive phrases are treated as latent variables is proposed to deal with motion ambiguity and partial occlusion in interactions.
- 126
- Scene-Aware Context Reasoning for Unsupervised Abnormal Event Detection in Videos124
A scene-aware context reasoning method that exploits context information from visual features for unsupervised abnormal event detection in videos, which bridges the semantic gap between visual context and the meaning of abnormal events is proposed.
- Overcoming Language Priors in VQA via Decomposed Linguistic Representations120
A novel method of language attention-based VQA that learns decomposed linguistic representations of questions and utilizes the representations to infer answers for overcoming language priors is presented.
- A Hyperbolic-to-Hyperbolic Graph Convolutional Network111
A manifold-preserving graph convolution that consists of ahyperbolic feature transformation and a hyperbolic neighborhood aggregation and achieves substantial improvements on the link prediction, node classification, and graph classification tasks.
- Discriminative human action recognition in the learned hierarchical manifold space104
This paper proposes a hierarchical discriminative approach for human action recognition that consists of feature extraction with mutual motion pattern analysis and discrim inative action modeling in the hierarchical manifold space.
- NON-NEGATIVE MATRIX FACTORIZATION FRAMEWORK FOR FACE RECOGNITION103
A framework of face recognition by adding NMF constraint and classifier constraints to matrix factorization to get both intuitive features and good recognition results is proposed and two novel subspace methods are presented.
- 99
- 97
- Intrinsic Image Decomposition Using Optimization and User Scribbles91
This paper presents a novel high-quality intrinsic image recovery approach using optimization and user scribbles that achieves a better recovery result of intrinsic reflectance and illumination components than the previous approaches.
- 88
- Segmentation of the left ventricle in cardiac cine MRI using a shape-constrained snake model78
This paper presents an automatic LV myocardial boundary segmentation method using the parametric active contour model (or snake model), and a novel shape-similarity based energy is proposed to prevent the snake contour from being strapped in faulty edges and to preserve weak boundaries.
- 77
- A deep Coarse-to-Fine network for head pose estimation from synthetic data76
This paper designs a rendering pipeline to synthesize realistic head images and generates an annotated dataset with a collection of 310k head poses, and proposes a deep neural network following the Coarse-to-Fine strategy to estimate head poses.
- Temporal dynamic appearance modeling for online multi-person tracking75
This work proposes a temporal dynamic appearance model for online multi-person tracking, and presents an incremental learning approach to learn the model online, which outperforms the state-of-the-art multi- person tracking algorithms.
- 72
- View-Invariant Action Recognition Using Latent Kernelized Structural SVM70
A novel learning algorithm is proposed for the view-invariant action recognition, which extends the kernelized structural SVM framework to include latent variables and combines the low-level visual cue, mid-level correlation description, and high-level label information into a novel nonlinear kernel.
- A Decomposition Model for Stereo Matching69
A decomposition model for stereo matching is presented to solve the problem of excessive growth in computational cost as the resolution increases and achieves 10−100× speed increase while obtaining comparable disparity estimation results.
- 69
- Deep 3D Portrait From a Single Image68
Extensive evaluations and comparison with previous methods show that the new learning-based approach for recovering the 3D geometry of human head from a single portrait image can produce high-fidelity 3D head geometry and head pose manipulation results.
- Deep Stereo Matching With Explicit Cost Aggregation Sub-Architecture68
A learning-based cost aggregation method for stereo matching by a novel sub-architecture in the end-to-end trainable pipeline that outperforms the state-of-the-art methods.
- A linear discriminant analysis framework based on random subspace for face recognition68
A novel framework, random discriminant analysis (RDA), is proposed to handle the small sample size problem when dealing with high-dimensional face data and can take full advantage of useful discriminant information in the face space.
- 66
- Multi-modal Agent Tuning: Building a VLM-Driven Agent for Efficient Tool Usage65
A multi-modal agent tuning method that automatically generates multi-modal tool-usage data and tunes a vision-language model (VLM) as the controller for powerful tool-usage reasoning is proposed.
- 65
- Deep convolutional network with locality and sparsity constraints for texture classification61
This paper advocates exploiting appropriately convolutional layer activations to constitute a powerful descriptor for texture classification under an end-to-end learning framework that notably outperforms the state-of-the-art methods.
- Optimal Essential Matrix Estimation via Inlier-Set Maximization58
The globally optimal “rotation space search” method is extended to essential matrix estimation in the presence of feature mismatches or outliers, finding an explicit, geometrically meaningful essential manifold parametrization using a 5D direct product space of a solid 2D disk and a solid 3D ball.
- 58
- 55
- Joint Commonsense and Relation Reasoning for Image and Video Captioning53
A joint commonsense and relation reasoning method that exploits prior knowledge for image and video captioning without relying on any detectors, and experiments validate the effectiveness of the prior knowledge-based approach.
- 53
- 52
- Robust Match Fusion Using Optimization48
This algorithm is the first patch-based exposure fusion work to preserve the moving objects of dynamic scenes that does not need the registration process of different exposure images, and it achieves visually pleasing fusion results without ghosting artifacts.
- 46
- Gaussian mixture modeling and learning of neighboring characters for multilingual text extraction in images46
A discriminative training algorithm based on the maximum-minimum similarity (MMS) criterion to estimate the parameters in the proposed text extraction approach shows significant improvement of overall performance, compared with influential training criterions of the maximum likelihood (ML) and the maximum classification error (MCE).
- 44
- 40
- Revisiting Bilinear Pooling: A Coding Perspective39
A new feature fusion algorithm, the factorized bilinear coding (FBC) method, is devised, which can generate compact and discriminative representations with substantially fewer parameters.
- 39
- 38
- 38
- Learning a robust representation via a deep network on symmetric positive definite manifolds38
This paper proposes to aggregate deep convolutional features into an SPD matrix representation through the SPD generation and the SPD transformation under an end-to-end deep network, and shows that the method notably outperforms state-of-the-art methods on visual classification.
- 38
- 38
- Visual-Semantic Graph Matching for Visual Grounding37
This paper forms visual grounding as a graph matching problem to find node correspondences between a visual scene graph and a language scene graph, and learns unified contextual node representations of the two graphs by using a cross-modal graph convolutional network to reduce their discrepancy.
- 36
- 34
- Deep CNN based binary hash video representations for face retrieval34
A novel deep convolutional neural network is proposed to learn discriminative binary hash video representations for face retrieval, which integrates face feature extractor and hash functions into a unified optimization framework to make the two components be as compatible as possible.
- 32
- 32
- Incremental discriminant-analysis of canonical correlations for action recognition32
This paper proposes an online recognition method, namely incremental discriminant-analysis of canonical correlations (IDCC), in which the discriminative model is incrementally updated to capture the changes of human appearance, and thereby facilitates the recognition task in changing environments.
- Learning online structural appearance model for robust object tracking31
Both qualitative and quantitative evaluations on various challenging image sequences demonstrate that the proposed algorithm outperforms the state-of-the-art methods.
- 30
- 30
- 29
- Nonnegative correlation coding for image classification28
Based on the Nesterov’s gradient projection algorithm, an effective numerical solver is developed to optimize the nonnegative correlation coding problem with guaranteed quadratic convergence.
- A study of regularized Gaussian classifier in high-dimension small sample set case based on MDL principle with application to spectrum recognition27
This paper investigates the covariance matrix estimation problem for small-number samples with high dimension setting based on minimum description length (MDL) principle, and develops a new covariance Matrix estimator, and a formula for fast estimation of regularization parameters is derived.
- 26
- 26
- Exploring Data Geometry for Continual Learning25
This method dynamically expands the geometry of the underlying space to match growing geometric structures induced by new data, and pre-vents forgetting by keeping geometric structures of old data into account, and achieves better performance than baseline methods designed in Euclidean space.
- Evidential Reasoning for Video Anomaly Detection25
A deep evidential reasoning method that explicitly learns the uncertainty to model the boundaries of video anomalies, and introduces an unsupervised strategy to train the network by minimizing an energy function of the deep Gaussian mixed model (GMM).
- 24
- Adaptive learning codebook for action recognition24
The weighted adaptive metric learning (WAML) is proposed which integrates the two independent phases into a unified optimization framework and enables to select indispensable and crucial dimensions for building a discriminative codebook.
- Efficient Multi-turn RL for GUI Agents via Decoupled Training and Adaptive Data Curation23
This work proposes DART, a Decoupled Agentic RL Training framework for GUI agents, which coordinates heterogeneous modules in a highly decoupled manner and will fully open-source the training framework, data, and model checkpoints via computer-use-agents.
- 23
- Iterative Tool Usage Exploration for Multimodal Agents via Step-wise Preference Tuning21
This paper proposes an iterative tool usage exploration method for multimodal agents without any pre-collected data, namely SPORT, via step-wise preference optimization to refine the trajectories of tool usage to eliminate the bottleneck of human annotation.
- Recognising human interaction from videos by a discriminative model21
A novel discriminative method is proposed, which model the action of each person by a large-scale global feature and local body part features, to capture such interdependencies for recognising interaction of two people.
- Discriminative structure selection method of Gaussian Mixture Models with its application to handwritten digit recognition21
The proposed discriminative method to select GMM structures for pattern classification behaves better than the manual method and the generative counterparts, including Bayesian Information Criterion, Minimum Description Length (MDL) and AutoClass.
- Learning Event-Relevant Factors for Video Anomaly Detection20
A causal generative model is introduced to separate the event-relevant factors and event-irrelevant ones in videos, and the prototypes of event- relevant factors in a memory augmentation module are learned to eliminate the interferences from event-IRrelevant factors on anomaly predictions.
- 20
- External Force for Active Contours: Gradient Vector Convolution20
This paper presents a novel external force called gradient vector convolution (GVC) for active contours, motivated by gradient vector flow (GVF) and possesses some advantages of GVF, such as enlarged capture range, initialization insensitivity and high performance on concavity convergence.
- 20
- A Miniature Stereo Vision Machine for Real-Time Dense Depth Mapping20
A miniature stereo vision machine (MSVM-2) to generate high-resolution dense depth map for application to portable intelligent robots and smart visual interface and the whole algorithm is compactly implemented in a single FPGA.
- 19
- 19
- 18
- An Eye-Tracking Study of User Behavior in Web Image Search18
It was found that search results at certain locations, e.g., the top-center area in a grid layout, were more attractive than others and the search task type significantly influenced user behavior while the results presentation order didn’t.
- 18
- A Hierarchical Video Description for Complex Activity Understanding17
This paper develops a latent discriminative structural model to detect the complex activity and atomic actions, while learning the temporal structure of atomic actions simultaneously, and presents a semi-supervised learning method to automatically predict the atomic action labels of unlabeled training videos when the labeled training data is limited.
- Background modeling by subspace learning on spatio-temporal patches17
This paper presents a novel background model for video surveillance-Spatio-Temporal Patch based Background Modeling (STPBM), which uses spatio-temporal patches, called bricks, to characterize both the appearance and motion information.
- 17
- Face recognition with local steerable phase feature17
A novel local steerable phase (LSP) feature extracted from the face image using steerable filters for face recognition is proposed which is semi-invariant under common image deformations and distinctive enough to provide useful identity information.
- 16
- A low-cost tele-presence wheelchair system16
The user interface developed in the lab allows an operator to access the system anywhere and directly touch the live video image of the wheelchair to push it as if he/she did it in the presence.
- 16
- An FPGA-based RGBD imager16
A fine-grain parallel and pipelining architecture in FPGA is designed for implementation to achieve a high computational and real-time throughput and to reduce the hardware resource consumption and accelerate the processing speed of a trinocular stereo vision system.
- 16
- 15
- Stitching Videos from a Fisheye Lens Camera and a Wide-Angle Lens Camera for Telepresence Robots15
A multi-homography-based video stitching method which stitches videos from a wide-angle camera and a fisheye camera to provide a compact and efficient user interface for telepresence robots with user-friendly interactive experiences.
- 15
- A bottom-up algorithm for finding principal curves with applications to image skeletonization15
A bottom-up strategy to construct a graph called a principal graph for representing a principal curve by initializing a set of vertices based on principal oriented points introduced by Delicado, and then constructs the principal graph from these vertices through a two-layer iteration process.
- 14
- 14
- On the Critical Point of Gradient Vector Flow Snake14
The so-called critical point problem of Gradient vector flow (GVF) snake is studied in two respects: influencing factors and detection of the critical points.
- 14
- 13
- 13
- Unsupervised Deep Learning of Mid-Level Video Representation for Action Recognition13
This paper proposes an unsupervised deep learning method which employs unlabeled local spatial-temporal volumes extracted from action videos to learn midlevel video representation for action recognition, and simultaneously discovers mid-level semantic concepts by discriminative clustering and optimizes local temporal features by two relatively small and simple deep neural networks.
- Orthonormal dictionary learning and its application to face recognition13
The ability of the class-specific dictionary to well represent samples from the associated class and suppress the ability of representing samples from other classes is enhanced, and the representations that have small within- class scatter and big between-class scatter are enforced.
- Heterogeneous discriminant analysis for cross-view action recognition13
An approach of cross-view action recognition, in which the samples from different views are represented by features with different dimensions, is proposed, Inspired by linear discriminant analysis (LDA), and a discriminative common feature space is introduced to bridge the source and target views.
- Learning a discriminative mid-level feature for action recognition13
Experiments on the Weizmann, UCF sports, Ballet, and multi-view IXMAS datasets demonstrate that out mid-level feature learned from multiple low-level features can achieve a superior performance over state-of-the-art methods.
- 13
- 13
- Diving into the Fusion of Monocular Priors for Generalized Stereo Matching12
A binary local ordering map is proposed to guide the fusion, which converts the depth map into a binary relative format, unifying the relative and absolute depth representation and formulate the final direct fusion of monocular depth to the disparity as a registration problem.
- Memory-Centric Embodied Question Answering12
MemoryEQA, an EQA framework centered on memory, which establishes mechanisms for memory storage, update, and retrieval, allowing memory information to contribute throughout the entire exploration process, underscores the memory capability's pivotal role in solving complex tasks.
- 12
- 12
- 12
- A Fusion Framework of Stereo Vision and Kinect for High-Quality Dense Depth Maps12
A global energy function with a novel data term is designed, which provides a reasonable, straight-forward and scalable way to fuse stereo vision and the depth data from Kinect.
- 11
- In-Context Compositional Generalization for Large Vision-Language Models11
A demonstration selection method to achieve ICCG for LVLMs is proposed, by considering two key factors of demonstrations: content and structure, from a multimodal perspective, and design a diversity-coverage-based matching score to select demonstrations with maximum coverage.
- 11
- 11
- 11
- 11
- Combining evolution strategy and gradient descent method for discriminative learning of bayesian classifiers11
A hybrid approach of the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) and the gradient decent method for optimizing Bayesian classifiers under the SOFT target based Max-Min posterior Pseudo-probabilities (Soft-MMP) learning framework is proposed.
- Discriminant Clustering Embedding for Face Recognition with Image Sets11
A novel local discriminant embedding method, Discriminant Clustering Embedding (DCE), is proposed for face recognition with image sets that combines the effectiveness of submanifolds, which are extracted by clustering for each subject's image set, and the inherent structure of face appearance manifold.
- Segmentation of the Left Venctricle from MR Images via Snake Models Incorporating Shape Similarities11
- Residual Hyperbolic Graph Convolution Networks10
This paper introduces a hyperbolic residual connection function to overcome the over-smoothing problem, and theoretically proves the effectiveness of the hyperbolic residual function.
- Heterogeneous domain adaptation method for video annotation10
A novel feature learning method, called heterogeneous discriminative analysis of canonical correlation (HDCC), is proposed to discover a common feature subspace in which heterogeneous features can be compared, and a group weighting learning framework for multi-domain adaptation to effectively leverage knowledge learned from the source domain.
- 10
- 10
- 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.
- 9
- Cardiac Motion Estimation from Tagged MRI Using 3D-HARP and NURBS Volumetric Model9
Experimental results on in vivo data show that the proposed strategy could estimate 3D motion of LV rapidly and effectively benefiting from both HARP and NURBS model.
- 9
- 9
- Fine-Grained Annotation for Face Anti-Spoofing8
A fine-grained annotation method for face anti-spoofing by leveraging the Segment Anything Model to obtain pixel-wise segmentation masks by utilizing face landmarks as point prompts and introducing the Multi-Channel Region Exchange Augmentation to diversify training data and reduce overfitting.
- Learning a Gradient-free Riemannian Optimizer on Tangent Spaces8
A simple yet efficient Riemannian meta-optimization method that learns to optimize on tangent spaces of manifolds is proposed, which brings efficient learning of the optimizer, while enjoying a good optimization trajectory in a data-driven manner.
- 8
- Structured-Patch Optimization for Dense Correspondence8
A new method to compute the dense correspondences between two images by using the energy optimization and the structured patches using the property of the sparse feature and the principle that nearest sub-scenes and neighbors are much more similar is presented.
- Multi-task l0 gradient minimization for visual tracking8
A visual tracking method is proposed based upon l0-norm minimization which directly seeks solution to the primal l0 problem to avoid solving a large number of l0 minimization problems.
- 8
- Tracking pointing gesture in 3D space for wearable visual interfaces8
A wearable stereo vision system for pointing gesture tracking, called POGEST, with FPGA-based dense depth mapping at video rate, and a strategy of fusing information from selective attention maps and synthetical feature maps for locating the focus of attention pointed at by pointing gesture.
- 8
- 7
- 7
- Stitching images from a conventional camera and a fisheye camera based on nonrigid warping7
This paper proposes to stitch images from conventional cameras and fisheye cameras for offering remote operators a large field of view to perceive a local environment and exploits a region-based point correspondence selection method to reduce the number of point correspondences that are used for thin-plate spline interpolation.
- 7
- Shape Constraints for the Left Ventricle Segmentation from Cardiac Cine MRI Based on Snake Models7
This chapter addresses the shape constraints for extracting the endocardium and epicardium from cardiac cine MRI based on snake models and proposes a novel shape similarity energy for epicardia segmentation.
- 7
- Tracking articulated objects by learning intrinsic structure of motion7
A novel dimensionality reduction method, temporal neighbor preserving embedding (TNPE), to learn the low-dimensional intrinsic motion manifold of articulated objects by combining Bayesian mixture of experts with Gaussian mixture model to establish a probabilistic non-linear mapping from the embedding space to the configuration space.
- 7
- EyeScreen: A Gesture Interface for Manipulating On-Screen Objects7
A gesture-based interaction system which provides a natural way of manipulating on-screen objects and a method based on projective geometry for background subtraction to improve the system performance is presented.
- Learning the Dynamics of Visual Relational Reasoning via Reinforced Path Routing6
A reinforced path routing method that represents an input image via a structured visual graph and introduces a reinforcement learning based model to explore paths (sequences of nodes) over the graph based on an input sentence to infer reasoning results.
- 6
- Online maximum a posteriori tracking of multiple objects using sequential trajectory prior6
This paper introduces the sequential trajectory prior, i.e., the prior information from previous frames about “good” trajectories, into the two MAP stages, which jointly optimizes the detections and trajectories to facilitate online multi-object tracking.
- Can You Easily Perceive the Local Environment? A User Interface with One Stitched Live Video for Mobile Robotic Telepresence Systems6
The results show that the user interface with one stitched live video improves task efficiency, the number of errors, and remote operators’ feelings of presence, and enables remote operators to concentrate on the work they are doing.
- 6
- Recognizing key segments of videos for video annotation by learning from web image sets6
An adaptive latent structural SVM method to adapt the pre-learned classifiers using Web image sets to an optimal target classifier, where the locations of the key segments are modeled as latent variables because the ground-truth of key segments is not available.
- Inferring User Preference in Good Abandonment from Eye Movements6
This paper uses eye movement data to infer which search result has satisfied user’s information need in each good abandonment instance, and shows that the approach can infer user preference with reasonable accuracy.
- Online visual tracking by integrating spatio‐temporal cues6
This study proposes a novel online tracking algorithm by integrating spatio-temporal cues to alleviate the drift problem and develops an efficient cue quality evaluation criterion that combines appearance and motion information.
- 6
- A Lateral Inhibitory Spiking Neural Network for Sparse Representation in Visual Cortex6
This paper finds that lateral inhibition is the fundamental neural mechanism for sparse representation in the visual cortex, by which cortical neurons not only compete with eachother so that the input signal can be represented sparsely but also cooperate with each other to make the representation more accurate.
- 6
- 6
- 6
- Curvature Learning for Generalization of Hyperbolic Neural Networks5
A PAC-Bayesian generalization bound of HNNs is derived, highlighting the role of curvatures in the generalization of HNNs via their effect on the smoothness of the loss landscape, and a sharpness-aware curvature learning method is proposed to smooth the loss landscape, thereby improving the generalization of HNNs.
- Efficient Riemannian Meta-Optimization by Implicit Differentiation5
An efficient Riemannian meta-optimization method that decouples the complex computation scheme from the meta-gradient by establishing a link between RiemANNian optimization and the implicit function theorem.
- Compact discriminative object representation via weakly supervised learning for real‐time visual tracking5
The authors propose a compact discriminative object representation to achieve both good tracking accuracy and efficiency and demonstrate that the proposed algorithm is much more computationally efficient than the state-of-the-art methods with comparable accuracy.
- 5
- 5
- Differential tracking with a kernel-based region covariance descriptor5
This paper presents a kernel-based region covariance descriptor that is incorporated to the covariance matrix to effectively control the correlations among extracted features inside the object region.
- Transfer Discriminant-Analysis of Canonical Correlations for View-Transfer Action Recognition5
A novel transfer learning approach to recognize human actions from one view (target view) via the discriminative model learned from another view, which proposes a Weighted Canonical Correlations scheme to fuse the multi-class canonical correlations from multiple source views according to their corresponding weights for recognition in the target view.
- 5
- 5
- Character Stroke Extraction Based on B-spline Curve Matching by Constrained Alternating Optimization5
- Wavelat-Based Resolution Enhancement of Omnidirectional Images5
The experimental results demonstrate the proposed approach is feasible and efficient for resolution enhancement of omindirectional image and a fine resolution enhancement image reconstructed via inverse wavelet transform is discussed.
- Multi-Label Stereo Matching for Transparent Scene Depth Estimation4
This paper presents a multi-label stereo matching method to simultaneously estimate the depth of the transparent objects and the occluded background in transparent scenes and introduces a pixel-wise multivariate Gaussian representation to resolve the multi-label regression problem.
- Unsupervised deep quantization for object instance search4
The proposed unsupervised deep quantization method for object instance search is optimized with an iterative optimization strategy which guarantees that the features and the supervision information can be enhanced each other alternately in a unified model.
- Vehicle Verification Based on Deep Siamese Network with Similarity Metric4
An end-to-end model based on the Siamese Convolutional Neural Network (CNN), which integrates distance metric learning and feature learning into a unified framework for vehicle verification is proposed.
- A discriminative structural model for joint segmentation and recognition of human actions4
A novel discriminative structural model is proposed for splitting a long-term video into segments and annotating the action label of each segment to exploit the statistical dependences among segments.
- Discriminative Orthonormal Dictionary Learning for Fast Low-Rank Representation4
With the learned discriminative orthonormal dictionary learning method, the low-rank representation problem can be solved much faster than traditional methods.
- Robust Online Multi-object Tracking by Maximum a Posteriori Estimation with Sequential Trajectory Prior4
This paper estimates the optimal object trajectories by solving two MAP estimation problems: object detection and trajectory-detection association by introducing the sequential trajectory prior into MAP estimation, which allows the two MAP stages interact with each other in a sequential manner, which facilitates online multi-object tracking.
- 4
- Automatic Image Annotation with Cooperation of Concept-Specific and Universal Visual Vocabularies4
An automatic image annotation method based on concept-specific image representation and discriminative learning of Bayesian classifiers, Max-Min posterior Pseudo-probabilities (MMP), which is promising with comparisons to state-of-the-art results.
- 4
- 4
- 4
- Maximum-Minimum Similarity Training for Text Extraction4
This paper uses the discriminative training criterion of maximum-minimum similarity (MMS) to improve the performance of text extraction based on Gaussian mixture modeling of neighbor characters and defines the corresponding objective function for text extraction.
- Modality Alignment across Trees on Heterogeneous Hyperbolic Manifolds3
A semantic-aware visual feature extraction framework that applies a cross-attention mechanism to visual class tokens from intermediate Transformer layers, guided by textual cues to extract visual features with coarse-to-fine semantics is introduced.
- Adversarial Sample Synthesis for Visual Question Answering3
This article proposes an adversarial sample synthesis method, which synthesizes different adversarial samples by adversarial masking at different training epochs to cope with the changing memorized language priors of VQA models.
- 3
- A Tele-Presence Wheelchair for Elderly People.3
The architecture and implementation of a tele-Presence wheelchair based on tele-presence robot, intelligent wheelchair, and smart pad TIUI technologies is presented and the preliminary evaluation of the user experience is discussed.
- Intercell scheduling: A negotiation approach using multi-agent coalitions3
Computational results show that the coalition-based scheduling (CBS) approach using distributed multi-agent negotiation is better than the existing intercell scheduling approach for the same problem with respect to the solution quality and computational costs.
- Attention Estimation for Input Switch in Scalable Multi-display Environments3
A method of automatic switch for routing one input to different displays in scalable MDEs based on the user attention estimation by using Convolutional Neural Networks to learn the attention model from face videos with various poses, illuminations, and occlusions.
- Completed Dense Scene Flow in RGB-D Space3
A completed dense scene flow framework is proposed, which models both rotation and translation for general motion estimation and outperforms other published methods on the RGB-D image space.
- Video Annotation by Incremental Learning from Grouped Heterogeneous Sources3
An incremental transfer learning framework is proposed, in which heterogeneous knowledge are integrated and incrementally added to update the target classifier during learning process, which demonstrates the effectiveness of the proposed method on the two challenging video datasets.
- Landmark-Based Inductive Model for Robust Discriminative Tracking3
Qualitative and quantitative evaluations on \(65\) challenging image sequences including the benchmark dataset and other public sequences demonstrate that the proposed algorithm outperforms the state-of-the-art methods.
- Coupling Semi-supervised Learning and Example Selection for Online Object Tracking3
This paper introduces an active example selection stage between sampling and labeling, and proposes a novel online object tracking algorithm which explicitly couples the objectives of semi-supervised learning and example selection.
- Dynamic Construction of Multilayer Neural Networks for Classification3
The main goal of this work is to train a hidden neuron and assemble it to the network dynamically while making the learning error smaller and smaller, to successfully avoid the overlearning problem.
- Convolutional Virtual Electric Field External Force for Active Contours3
A novel external force called CONVEF for active contours is proposed by taking the Virtual Electric Field just as a convolution operation and by using modified distance metrics in the convolution kernel.
- 3
- Bits allocation of matching pursuit video coder based on visual properties3
Original atoms searching scheme will be modified in order to distribute the atoms around the interesting regions, thus the reconstructed images' quality of the matching pursuit video coder can be promoted.
- 3
- A Vision-Based Method for Finger-Screen Interaction3
Experimental results demonstrate that the interactive method can realize robust track- ing of finger motion, against complex background and varying illumination conditions, as well as the instant detection of finger click- ing actions.
- 2
- Hyperbolic Dual Feature Augmentation for Open-Environment2
A hyperbolic dual feature augmentation method for open-environment, which augments features for both seen and unseen classes in the hyperbolic space, and derives an upper bound for the hyperbolic dual augmentation loss.
- A method of embedding a high-resolution image into a large field-of-view image2
The results indicate that the proposed method can effectively embed a high-resolution image into a large FOV image as well as preserve the shape and content of the large FOV image, indicating that the resolution of a large FOV image can be improved via high-resolution image embedding.
- Learning Weighted Video Segments for Temporal Action Localization2
A Supervised Temporal Attention Network (STAN) is built to dynamically learn the weights of video segments, and generate descriptive and discriminative video representations, and achieves substantially better performance than the state-of-the-art methods.
- Fusing Appearance Features and Correlation Features for Face Video Retrieval2
This paper fuse appearance features and correlation features to exploit rich information of face videos for face video retrieval via a deep convolutional neural network, and integrates feature extractions, feature fusion, and hash learning into a unified optimization framework to guarantee optimal Compatibility.
- 2
- Online-Learning Structural Appearance Model for Robust Visual Tracking2
A robust tracking method with an online-learning structural appearance model based on local sparse coding and online metric learning that outperforms the state-of-the-art methods.
- Multiview Visibility Estimation for Image-Based Modeling2
This paper investigates the problem of determining regions in 3D scene visible to some given viewpoints when obstacles are present in the scene in an implicit framework and applies the multiview visibility estimation to an image-based modeling technique.
- 2
- 2
- A multimodal labeling interface for wearable computing2
In this paper, a concept of virtual touchpad based on stereo vision techniques is proposed, and the touch events and movements of a pointing gesture are robustly detected for natural gesture interactions.
- Action recognition feedback-based framework for human pose reconstruction from monocular images2
A novel framework based on action recognition feedback for pose reconstruction of articulated human body from monocular images is proposed and it is demonstrated that this method performs better on estimation accuracy than other methods withoutaction recognition feedback.
- 2
- 2
- 2
- Linear Self-Calibration Technique Based on Active Vision2
Simulative and real experimental results show that the intrinsic parameters of the camera's intrinsic parameters can be calculated precisely by applying the linear self-calibration technique.
- 2
- A Morphological Algorithm for Color Image Multi-scale Segmentation2
A multi-scale morphological algorithm for color images segmentation is proposed in this paper and a RAG and a NNG are constructed to accelerate following morphological processing.
- 1
- 1
- Long-Horizon Visual Imitation Learning via Plan and Code Reflection1
A new agent framework is proposed that incorporates two dedicated reflection modules that jointly enable the agent to detect and correct errors in both the plan generation and code generation, improving performance in tasks with intricate temporal and spatial dependencies.
- Large-scale Riemannian meta-optimization via subspace adaptation1
This paper proposes an efficient Riemannian meta-optimization method that significantly reduces the memory burden for large-scale optimization via a subspace adaptation scheme and can not only reduce the memory consumption but also improve the performance of Riemannian meta-optimization.
- 1
- 1
- 1
- 1
- 1
- 1
- Tracker-Level Decision by Deep Reinforcement Learning for Robust Visual Tracking1
This paper formulate the multi-tracker tracking problem as a decision-making task and train an expert by the deep reinforcement learning (DRL) to select the best tracker to ensure the reliability of the tracker.
- Incremental transfer learning for video annotation via grouped heterogeneous sources1
An incremental transfer learning framework is presented to integrate heterogeneous source knowledge and update the annotation model incrementally during the transfer learning process to reduce the tedious and expensive labelling efforts required in video annotation.
- Temporal Invariant Factor Disentangled Model for Representation Learning1
A disentangled model is proposed to disentangle from an image sequence the temporal-invariant factor that is used as an object representation insensitive to appearance changes and is built upon the variational auto-encoder and the recurrent neural network to independently approximate the posterior distributions of the factor in an unsupervised manner.
- Set-to-Set Distance Metric Learning on SPD Manifolds1
This paper proposes a novel SPD matrix metric learning method of discovering SPD matrix intrinsic properties and measuring the distance considering different roles of intrinsic properties, which achieves the state-of-the art performance.
- Online Multi-Person Tracking Based on Metric Learning1
A discriminative appearance model is constructed by using metric learning which can obtain accurate appearance affinities with human appearance variations and can significantly guide data association.
- 1
- Hebbian-based mean shift for learning the diverse shapes of V1 simple cell receptive fields1
A biologically plausible learning algorithm, named Hebbian-based mean shift, is presented for sparse coding that optimizes the number of basis functions rather than their coefficients, and can be interpreted as mode detection where the basis functions are the modes of the kernel density estimate.
- Text-Independent Phoneme Segmentation via Learning Critical Acoustic Change Points1
This paper presents a learning-based method in which the phoneme boundaries are viewed as critical points in the acoustic change context of speech signals, and adopts a metric learning procedure in the calculation of acoustic changes, in order to make the acoustic changes at phoneme boundary more discriminative.
- 1
- 1
- 1
- Soft Measure of Visual Token Occurrences for Object Categorization1
A soft bag-of-features image representation based on Gaussian Mixture Modeling of visual tokens for object categorization under a new discriminative learning framework of Bayesian classifiers, the Max-Min posterior Pseudo-probabilities (MMP).
- Stereo Vision Based Floor Plane Extraction and Camera Pose Estimation1
A histogram based method is performed on the reconstructed 3-D points obtained by an onboard video-rate stereo vision system to extract candidate points of the floor plane and determine the pose of vision system simultaneously.
- Surface Reconstruction from Images Using a Variational Formulation1
This paper presents a new approach to recovering the surface of a 3D object from multiple calibrated images based on a variational formulation which defines an energy functional where both silhouette and texture consistency constraints are included.
- Text Extraction Based on Maximum-Minimum Similarity Training Method1
A maximum-minimum similarity training algorithm to optimize the parameters in the effective method of text extraction based on Gaussian mixture modeling of neighbor characters and uses the gradient descent method to search the minimum of the objective function and the optimum parameters for the text extraction method.
- 1
- EyeScreen:A Vision-Based Gesture Interaction System1
EyeScreen is a vision-based interaction system which provides a natural gesture interface for human-computer interaction by tracking human fingers and recognizing gestures and enables users to directly interact with rendered objects on the screen.
- 1
- An Adaptable Failure Detection Method for High Reliable Distributed Computing Systems1
An adaptable heartbeat failure detection method that can adapt the change of the state of high reliable distributed computing system, achieve a compromise between a good detection time and the need of avoiding false detections.
- Face Recognition Based on Local Steerable Feature and Random Subspace LDA1
This paper proposes a novel local steerable feature extracted from the face image using steerable filter for face representation that is locally stable with respect to scale, noise and brightness changes and it is semi-invariant under common image deformations and distinctive enough to provide useful identity information.
- Visual Hand Tracking Using Nonparametric Sequential Belief Propagation1
A statistical method for robust visual hand tracking, in which graphical model decoupling different hand joints is performed to represent the hand constraints, which is robust for 3D hand motion tracking.
- Camera Self-Calibration with Skew Factor1
Experimental results of synthetic data as well as real image sequences show that this algorithm can estimate the camera's intrinsic parameters precisely and make the successive 3D reconstruction more easily.
- New E-Commerce Model Based on Multi-Agent Automated Negotiation1
An e-commerce model based on multi-agent automated negotiation model is developed and evaluated, in which two competitive agents have firm deadlines and incomplete information about each other and the optimal negotiation strategy is analyzed and presented.
- Practical algorithm of building Delaunay triangle mesh for terrain modeling1
A practical algorithm to reduce sampling quantity of terrain in a single delaunay triangle mesh and to increase the efficiency of building the mesh greatly is introduced.
- Image stabilization for object tracking1
This paper presents a motion model to estimate the motion parameters for image stabilization by using feature points extracted from consecutive frames, which shows that the stabilization rate is up to 25 frames per second with tracking window size 1 12x 128 pixels.
- Person real-time tracking for video communication1
To improve the performance of block- based tracking during person turning around, a cylindrical model is introduced to approximate head, and extract features in the warped cylindle surface to correct the results of block tracking.
- Game-Theoretic Modeling of Heterogeneous Investor Interactions for Stock Price Forecasting–
This work proposes a novel game-theoretic modeling method that captures heterogeneous investor interactions for stock price forecasting and effectively outperforms state-of-the-art stock price forecasting methods.
- –
- –
- –
- Neural 3D Scene Reconstruction from Multiple 2D Images without 3D Supervision–
This paper proposes a novel neural reconstruction method that reconstructs scenes without 3D supervision on the ScanNet dataset and achieves competitive reconstruction compared with some existing methods that use 3D information as supervision on the ScanNet dataset.
- –
- Heterogeneous Multi-group Adaptation for Event Recognition in Consumer Videos–
A heterogeneous multi-group adaptation method to partition loosely labeled Web images and videos into several semantic groups and find the optimal weight for each group and demonstrate the effectiveness of this method.
- –
- Multi‐group–multi‐class domain adaptation for event recognition–
The authors’ framework is extended from multi-class support vector machine by adding a novel data-dependent regulariser, which can force the event classifier to become consistent in consumer videos.
- Weakly Supervised Action Recognition and Localization Using Web Images–
A novel Transfer Latent Support Vector Machine (TLSVM) is developed by using Web images and weakly annotated training videos to alleviate the laborious and time-consuming manual annotations of action locations.
- –
- –
- Cross-domain structural model for video event annotation via web images–
This paper uses a new discriminative structural model called Cross-Domain Structural Model (CDSM) to transfer knowledge from Web images to consumer videos (target domain), by jointly modeling the interaction between videos and images.
- High Quality Intrinsic Images Using Optimization–
This paper presents a novel intrinsic image recovery approach using optimization based on the assumption of color characteristics in a local window in natural images, which achieves a better recovery of intrinsic reflectance and illumination components than by previous approaches.
- –
- –
- –
- Layer-Constraint-Based Visibility for Volumetric Multi-view Reconstruction–
A novel approach called layer-constraint-based visibility (LCBV) to estimating visibility, based on the layered state of a scene and photo-consistency constraint, which can determine the more accurate visibility for every point in a scene.
- Human Ear Detection and Recognition Based on Face Profile and Ear Feature–
A new method which combined face profile, ear feature and relative position of ear and face profile to detect and recognize human ear is proposed in this paper.
- –
- An Adaptive Failure Detection Method for Distributed Computing Systems–
Simulation results show that the failure detector can achieve a compromise between a good detection time and the need of avoiding false detections, and can improve the failure detection QoS of the distributed computing systems.
- Facial Expression Analysis on Semantic Neighborhood Preserving Embedding–
In this study, an expression manifold is constructed by Neighborhood Preserving Embedding based on the expression semantic metric for a global representation of all possible facial expression images.
- Atomic multicast protocol for distributed computing systems of planetary rover–
The experiment shows that the atomic multicast protocol provides an inexpensive way to achieve reliable group multicast and has the advantages of lightweight and fastness.
- PRECISE SHAPE MEASUREMENT OF DYNAMIC SURFACE VIA SINGLE CAMERA STEREO–
Under the framework of RANSAC, the curves are interpolated robustly with cubic spline based on moving-least-square (MLS) and principal curves theory is employed to extract curves in images for stereo matching.
- –
- Panoramic Mosaicing Based on Multi-domain Analysis and Global Optimization–
A novel framework of panoramic mosaicing is presented based on phase correlation, particle filter and intensity difference minimizing, combining the characteristics of frequency domain and spatial domain and a strategy to reduce the dimensions of parameter space is developed.
- Distributed Computing System Architecture for Planetary Rover–
This paper discusses the system hardware architecture, software architecture, system communication architecture with system control model, system fault tolerant model and system reliability for designing planetary rover.
- Linear Discriminant Analysis in Complementary Subspace for Face Recognition–
Based on random subspace,a complementary subspace linear discriminant analysis (LDA) approach is presented for face recognition that combines the two parts discriminative features on the feature level and can effectively solve the small sample size problem of LDA.
- –
- Hand Tracking and Gesture Recognition Method Based on Wearable Vision–
The proposed hand tracking and gesture recognition method is an extension of the Icondensation algorithm, which integrates the shape and depth information provided by a wearable vision system for robust hand tracking.
- Algorithm for Image Segmentation Based on Anisotropic Diffusion–
A new algorithm for image segmentation based on anisotropic diffusion based on the regularity condition of the decomposed equation is proposed and provides recognition and understanding processing with improved input.
- Simulated Annealing Based Hand Tracking in a Discrete Space–
An extended simulated annealing method with particle filter is proposed to search for optimal hand configuration in this discrete space, in which simplex search running in multi-processor is designed to predict the hand motion instead of initializing the simulatedAnnealing randomly, and particle Filter is employed to represent the state of the tracker at each layer for searching in high dimensional configuration space.
- Intention Tracking Based Reinforcement Learning Agent Model–
A reinforcement learning agent model with intention tracking has been proposed to overcome the lagging in action in dynamic confrontation multi-agent systems (MAS) environment and results prove that agents with this model are more autonomic and adaptive.
- A HIGH-RESOLUTION COMPLIANT TACTILE SENSOR–
A high-resolution compliant tactile sensor based on total internal reflection (TIR)technique is presented, suitable for the sen-sor incorporated in the finger of a robot gripper to efficiently acquire parameters about the shape, position and orientation of an object for the grasping and recognition tasks.
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 2026-10-15. 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.