Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks318 from public data
- 4,660
- Beyond Part Models: Person Retrieval with Refined Part Pooling (and A Strong Convolutional Baseline)2,522
This paper targets at learning discriminative part-informed features for person retrieval and lays emphasis on the content consistency within each part, resulting in refined parts with enhanced within-part consistency.
- Beyond Part Models: Person Retrieval with Refined Part Pooling.2,442
This paper targets at learning discriminative part-informed features for person retrieval and lays emphasis on the content consistency within each part, resulting in refined parts with enhanced within-part consistency.
- Towards Real-Time Multi-Object Tracking1,148
This work incorporates the appearance embedding model into a single-shot detector, such that the model can simultaneously output detections and the corresponding embeddings, and is formulated as a multi-task learning problem.
- MARS: A Video Benchmark for Large-Scale Person Re-Identification1,088
It is shown that CNN in classification mode can be trained from scratch using the consecutive bounding boxes of each identity, and the learned CNN embedding outperforms other competing methods considerably and has good generalization ability on other video re-id datasets upon fine-tuning.
- SVDNet for Pedestrian Retrieval825
This paper proposes the SVDNet for retrieval problems, with focus on the application of person re-identification (reID), and shows that RRI effectively reduces the correlation among the projection vectors, produces more discriminative FC descriptors, and significantly improves the re-ID accuracy.
- Person Re-identification by Video Ranking642
A novel model to automatically select the most discriminative video fragments from noisy image sequences of people where more reliable space-time features can be extracted, whilst simultaneously to learn a video ranking function for person re-id is presented.
- SegFlow: Joint Learning for Video Object Segmentation and Optical Flow446
This paper proposes an end-to-end trainable network, SegFlow, for simultaneously predicting pixel-wise object segmentation and optical flow in videos, and demonstrates that introducing optical flow improves the performance of segmentation, against the state-of-the-art algorithms.
- Perceive Where to Focus: Learning Visibility-Aware Part-Level Features for Partial Person Re-Identification388
A Visibility-aware Part Model (VPM) is proposed for partial re-ID, which learns to perceive the visibility of regions through self-supervision, which significantly improves the learned feature representation and the achieved accuracy is on par with the state of the art.
- 366
- 317
- Fast and Accurate Online Video Object Segmentation via Tracking Parts275
This paper proposes a fast and accurate video object segmentation algorithm that can immediately start the segmentation process once receiving the images, and performs favorably against state-of-the-art algorithms in accuracy on the DAVIS benchmark dataset, while achieving much faster runtime performance.
- Video Super-Resolution with Recurrent Structure-Detail Network269
A novel recurrent video super-resolution method which is both effective and efficient in exploiting previous frames to super-resolve the current frame by dividing the input into structure and detail components which are fed to a recurrent unit composed of several proposed two-stream structure-detail blocks.
- Person Re-Identification by Discriminative Selection in Video Ranking231
This work presents a novel model to automatically select the most discriminative video fragments from noisy/incomplete image sequences of people from which reliable space-time and appearance features can be computed, whilst simultaneously learning a video ranking function for person ReID.
- A survey of recent advances in visual feature detection229
The recent progress in visual feature detection is presented and future trends as well as challenges are identified and the relations among different kinds of features are covered.
- Packing and Padding: Coupled Multi-index for Accurate Image Retrieval223
A coupled Multi-Index (c-MI) framework to perform feature fusion at indexing level, which improves the retrieval accuracy significantly, while consuming only half of the query time compared to the baseline, and is well complementary to many prior techniques.
- 214
- Linkage Based Face Clustering via Graph Convolution Network200
This paper presents an accurate and scalable approach to the face clustering task, and shows that the proposed method does not need the number of clusters as prior, is aware of noises and outliers, and can be extended to a multi-view version for more accurate clustering accuracy.
- Video Super-Resolution With Temporal Group Attention199
This work proposes a novel method that can effectively incorporate temporal information in a hierarchical way and achieves favorable performance against state-of-the-art methods on several benchmark datasets.
- 152
- Learning Prompt-Enhanced Context Features for Weakly-Supervised Video Anomaly Detection145
A novel WS-VAD framework that focuses on efficient temporal modeling and anomaly innerclass discriminability, and a Prompt-Enhanced Learning (PEL) module that incorporates semantic priors using knowledge-based prompts to boost the discrimination of visual features while ensuring separability across anomaly subclasses.
- A2-FPN: Attention Aggregation based Feature Pyramid Network for Instance Segmentation145
Attention Aggregation based Feature Pyramid Network (A2-FPN) is proposed, to improve multi-scale feature learning through attention-guided feature aggregation and shows consistent gains on different instance segmentation frameworks.
- VisDrone-DET2018: The Vision Meets Drone Object Detection in Image Challenge Results133
A large-scale drone-based dataset, including 8, 599 images with rich annotations, including object bounding boxes, object categories, occlusion, truncation ratios, etc, is released, to narrow the gap between current object detection performance and the real-world requirements.
- 131
- Applying BERT to Document Retrieval with Birch129
This demonstration focuses on technical challenges in the integration of NLP and IR capabilities, along with the design rationale behind the approach to tightly-coupled integration between Python and Java Virtual Machine.
- Detecting Everything in the Open World: Towards Universal Object Detection128
UniDetector is proposed, a universal object detector that has the ability to recognize enormous categories in the open world, and surpasses the traditional supervised baselines by more than 4% on average without seeing any corresponding images.
- 121
- 116
- Revisiting Temporal Modeling for Video Super-resolution107
This work carefully study and compares three temporal modeling methods and proposes a novel Recurrent Residual Network (RRN) for efficient video super-resolution, where residual learning is utilized to stabilize the training of RNN and meanwhile to boost the super- resolution performance.
- CycAs: Self-supervised Cycle Association for Learning Re-identifiable Descriptions106
A different unsupervised method is introduced that allows us to learn pedestrian embeddings from raw videos, without resorting to pseudo labels, and adapt the discrete association process to a differentiable form, such that end-to-end training becomes feasible.
- Data-Uncertainty Guided Multi-Phase Learning for Semi-Supervised Object Detection101
This paper comprehensively considers divergent types of unlabeled images according to their difficulty levels, utilize them in different phases, and ensemble models from different phases together to generate ultimate results.
- Multi-Target Domain Adaptation with Collaborative Consistency Learning101
The proposed collaborative learning framework can effectively exploit rich structured information contained in both labeled source domain and multiple unlabeled target domains and performs favorably against state-of-the-art unsupervised domain adaptation methods specially trained on a single source-target pair.
- Feature representation for statistical-learning-based object detection: A review99
The feature sketch is drawn and new insights for feature utilization are provided in order to tackle future challenges of object detection through this review of feature representation in statistical learning based object detection.
- VL-Grasp: a 6-Dof Interactive Grasp Policy for Language-Oriented Objects in Cluttered Indoor Scenes92
This work proposes a novel interactive grasp policy, named Visual-Lingual- Grasp (VL-Grasp), to grasp the target specified by human language, and builds a new challenging visual grounding dataset to provide functional training data for robotic interactive perception in indoor environments.
- Accurate Image Search with Multi-Scale Contextual Evidences92
This paper proposes to integrate discriminative cues from multiple contextual levels, i.e., local, regional, and global, via probabilistic analysis, and shows that CNN feature is complementary to SIFT due to its semantic awareness and compares favorably to several other descriptors such as GIST, HSV, etc.
- 92
- 88
- 84
- 82
- Bayes Merging of Multiple Vocabularies for Scalable Image Retrieval82
This paper proposes a Bayes merging approach to down-weight the indexed features in the intersection set through explicitly modeling the correlation problem in a probabilistic view, and a joint similarity on both image- and feature-level is estimated for the indexed Features in the intersections set.
- Towards Discriminative Representation Learning for Unsupervised Person Re-identification81
This work proposes a cluster-wise contrastive learning algorithm (CCL) by iterative optimization of feature learning and cluster refinery to learn noise-tolerant representations in the unsupervised manner and adopts a progressive domain adaptation (PDA) strategy to gradually mitigate the domain gap between source and target data.
- 81
- 70
- Exploiting effective facial patches for robust gender recognition62
It is shown that the classification accuracy of the method is comparable with that of state-of-the-art systems, which validates the effectiveness of the proposed method.
- Unsupervised Visual Representation Learning by Graph-Based Consistent Constraints62
This paper proposes to use a cycle consistency criterion for mining positive pairs and geodesic distance in the graph for hard negative mining, and shows that the mined positive and negative image pairs can provide accurate supervisory signals for learning effective representations using Convolutional Neural Networks (CNNs).
- 60
- Intention Oriented Image Captions With Guiding Objects59
A novel approach for generating image captions with guiding objects (CGO), which constrains the model to involve a human-concerned object when the object is in the image and which outperforms the state-of-the-art models in the task with average F1 75.8.
- 59
- 58
- 58
- 56
- 55
- Spectral convolutional neural network chip for in-sensor edge computing of incoherent natural light54
The authors proposed integrated computing based on natural light by combining large-scale spectral filters on CMOS image sensor to facilitate vector-inner products from incoherent light, which is the first integrated optical computing utilizing natural light.
- 53
- 51
- 50
- 50
- 49
- Noisy Boundaries: Lemon or Lemonade for Semi-supervised Instance Segmentation?48
This paper constructs a framework for semi-supervised instance segmentation by assigning pixel-level pseudo labels, and proposes a noise-tolerant mask head by leveraging low-resolution features to combat the negative effects of noisy boundaries and enhance the positive impacts.
- 47
- 46
- 45
- 45
- 44
- 43
- 41
- 41
- 40
- Fine-residual VLAD for image retrieval40
This paper revisits the vector of locally aggregated descriptors (VLAD), which aggregates the residuals of local descriptors to their cluster centers, and proposes to generate a number of residual codebooks descended from the original clusters to generate the fine residuals.
- 39
- 38
- Softmax Dissection: Towards Understanding Intra- and Inter-Class Objective for Embedding Learning36
The softmax loss and its variants are widely used as objectives for embedding learning applications like face recognition and it is straightforward to tune each part to the best state with D-Softmax as objective, but the computation of the inter-class part is redundant and sampling-based variants of D- softmax are proposed to reduce the computation cost.
- 35
- 34
- 33
- 31
- 30
- Identity-Seeking Self-Supervised Representation Learning for Generalizable Person Re-identification29
This paper aims to learn a domain-generalizable person re-identification (ReID) representation from large-scale videos without any annotation, and proposes an Identity-seeking Self-supervised Representation learning (ISR) method that exhibits superior generalization ability.
- 29
- Violent Video Detection Based on Semantic Correspondence29
This paper proposes a novel violent video detection model based on semantic correspondence between audio-visual data from the same video that achieves quite competitive results on both public and self-built datasets.
- 28
- 28
- 26
- 26
- Reliability-Aware Prediction via Uncertainty Learning for Person Image Retrieval25
The Uncertainty-Aware Learning (UAL) method aims at providing reliability-aware predictions by considering data uncertainty and model uncertainty simultaneously, and shows superior performance on three challenging benchmarks under the vanilla single query settings.
- Incremental face clustering with optimal summary learning via graph convolutional network24
Compared with state-of-the-art static face clustering methods, the proposed approach can yield comparable accuracy while consuming much less time.
- 24
- 24
- Salient region detection via simple local and global contrast representation24
The experimental results show that the salient region detection method outperforms most state-of-the-art methods, reducing the mean absolute error by 41.74% and 28.57% compared to the previous best reported results on the MSRA-1000 and CSSD datasets, respectively.
- 24
- 23
- Eye/eyes tracking based on a unified deformable template and particle filtering23
This paper proposes a novel algorithm framework that can not only track the locations of eye/eyes accurately, but also obtain the eye contour parameters simultaneously.
- Fast Pedestrian Detection With Attention-Enhanced Multi-Scale RPN and Soft-Cascaded Decision Trees22
Inspired by the success of traditional pedestrian detectors, this work uses soft-cascaded decision trees instead of cascaded deep neural networks to achieve high accuracy and fast detection speed simultaneously and test the effectiveness of the proposed AEMS-RPN on several pedestrian detection datasets.
- 20
- 20
- 20
- 20
- Progressive Representation Adaptation for Weakly Supervised Object Localization18
This paper addresses the problem of weakly supervised object localization where only image-level annotations are available for training object detectors through progressive representation adaptation with two main steps: classification adaptation and detection adaptation.
- 18
- Face replacement with large-pose differences18
A 3D head model is proposed to be built from a single frontal face photo that can be rendered under arbitrary poses and illuminations, making it possible to do swapping for faces with large pose variations.
- 17
- 16
- 16
- EFENet: Reference-Based Video Super-Resolution with Enhanced Flow Estimation16
The novel flow refinement module of EFENet refines the flow regarding the furthest frame using all the estimated flows, which leverages the global temporal information within the sequence and therefore effectively reduces the alignment errors.
- Intra-Clip Aggregation For Video Person Re-Identification16
This paper proposes a novel video-based data augmentation scheme, termed as Synchronous Data Augmentation, which outperforms the most recent state-of-the-art methods and proposes a cascade integration module which hierarchically aggregates the intra-clip features with a linear-nonlinear combining projection.
- 16
- 16
- 16
- 15
- 15
- Contextual image search15
The key idea is to use the search-to-annotation technique and the contextual textual query mining scheme to determine the corresponding textual query, to finally get semantically similar search results.
- LIBA: Language Instructed Multi-granularity Bridge Assistant for 3D Visual Grounding14
This work proposes LIBA, a Language-Instructed multi-granularity Bridge Assistant tailored for 3D-VG task, which inherits reasoning prowess and world knowledge of LLM to bridge point clouds and texts at multiple granularities.
- 14
- 14
- 14
- 14
- Cross Teaching-Enhanced Multispectral Remote Sensing Object Detection With Transformer13
This article presents an object detection method for remote sensing imagery named multispectral detection transformer (multispectral DETR), a novel cross-teaching method between single-spectral and multispectral models, which alleviates the spectral interference issue caused by inconsistent target visibility.
- 13
- 13
- 13
- 13
- 13
- Cognitive pedestrian detector: Adapting detector to specific scene by transferring attributes13
This paper proposes a novel transfer learning framework for improving the performance of a generic detector by adapting the detector to a scene specific detector and outperforms the state-of-the-art adapting approaches in benchmark datasets.
- Dynamic Object Queries for Transformer-based Incremental Object Detection12
The Dynamic object Query-based DEtection TRansformer (DyQ-DETR), which incrementally expands the model representation ability to achieve stability-plasticity tradeoff, and presents the isolated bipartite matching for object queries in different phases, based on disentangled self-attention.
- 12
- 12
- 12
- 12
- Multi Features Combination for Pedestrian Detection12
A new approach for pedestrian detection in crowded scene from static images based on hybrid features, one type of middle-level features, which compose of multi features include gradient features, Edgelet features and haar-like features, three low-level feature sets.
- 11
- Visual reranking with improved image graph11
An improved reranking method for the Bag-of-Words (BoW) based image search, based on a directed image graph robust to outlier distraction, which shows that the rank-level feature fusion can be adopted in this re ranking method as well.
- 11
- Video object tracking using improved chamfer matching and condensation particle filter11
A novel video tracking algorithm based on edge information for gray videos based on the combination of a condensation particle filter and an improved chamfer matching is proposed that is stable, robust, and can effectively handle rotation distortion.
- 10
- 10
- R(Det)2: Randomized Decision Routing for Object Detection10
A novel approach to combine decision trees and deep neural networks in an end-to-end learning manner for object detection by proposing randomized decision routing with node selective and associative losses, which can boost the feature representative learning and network decision simultaneously.
- Semantic multimodal violence detection based on local-to-global embedding10
This paper argues that parsing abstract violence into semantics helps to obtain the essential representation of violence, and proposes a semantic multimodal violence detection framework based on local-to-global embedding.
- Parallel Cascade R-CNN for object detection in remote sensing imagery10
A novel detection architecture for remote sensing imagery to solve the problem of scale diversity on the basis of Cascade R-CNN is proposed and the detection performance of those categories with smaller objects has been improved.
- 10
- 9
- 9
- 9
- 9
- 9
- 9
- Deep representation learning for license plate recognition in low quality video images8
A novel backbone network structure called SACNN is designed, which combines convolutional neural network (CNN) and self-attention mechanism to learn non-linear representations for the structural patterns of characters in low quality video images.
- Comparison Study of Extraction Accuracy of 3D Facial Anatomical Landmarks Based on Non-Rigid Registration of Face Template8
There is no significant difference in the automatic landmark localization accuracy between the TH-OCR algorithm and the MeshMonk algorithm, and both can meet the needs of oral clinical applications to a certain extent.
- 8
- Dynamic temporal residual network for sequence modeling8
A dynamic temporal residual network (DTRN) is proposed by incorporating residual learning into an LSTM network along the temporal dimension and it is shown that the proposed DTRN has outperformed previously reported methods.
- 8
- Filter ensemble regularized common spatial pattern for EEG classification8
A filter ensemble technique for RCSP (FERCSP) to further extract frequency information and aggregate all the RCSPs efficiently to get an ensemble-based solution and the experimental results show that FERCSP significantly outperforms those of the existing methods in classification accuracy.
- 8
- Background Subtraction Based on a Combination of Local Texture and Color8
A new background subtraction algorithm is proposed in this paper, which combines local texture and color information to depict background and adopts the idea of mixture of Gaussian that uses multiple modes to represent background model.
- 7
- 7
- Face Quality Assessment via Semi-supervised Learning7
Experimental results show that the proposed end-to-end face quality assessment algorithm can improve the performance of face recognition by face selection and can get much higher accuracy than the models trained with the same amount of labeled faces.
- 7
- 7
- Human 3D Model-based 2D Action Recognition7
A new human 3D model-based 2D action recognition algorithm is presented in this paper for adapting to the actor s variable orientates, and the results prove robustness of the proposed method with respect to theActor s orientates and camera configurations.
- UniDetector: Towards Universal Object Detection With Heterogeneous Supervision6
The critical points for UniDetector are that it leverages images of multiple sources and heterogeneous label spaces in training through image-text alignment, which guarantees sufficient information for universal representations, and involves heterogeneous supervision training, which alleviates the dependence on the limited fully-labeled images.
- Diffusion Model Meets Non-Exemplar Class-Incremental Learning and Beyond6
This work proposes a simple, yet effective diffusion models to generate class-representative features highly similar to real features, which provides an effective way for exemplar-free knowledge memorization.
- Robust 3D face modeling and tracking from RGB-D images6
The method can generate accurate 3D face models from an RGB-D image or image sequence and robustly track the face even if the person is with large head rotations and various facial expressions.
- 6
- 5
- Consistency Mean-Teaching for Unsupervised Domain Adaptive Person Re-identification5
Experimental results show that CMT brings significant improvements over the baseline and achieves competitive accuracy on four popular UDA re-ID and is integrated into a single student-teacher framework and facilitate complementary benefits.
- Exploiting sparse representation in the P300 speller paradigm5
A novel method is proposed to cope with this problem through sparse representation for the P300 speller paradigm through sample selection and a channel-aware dictionary to reduce the effect of noise, which can improve performance and enhance the computing efficiency simultaneously.
- Reliability Evaluation of Public Security Face Recognition System Based on Continuous Bayesian Network5
This paper synthetically applies a variety of methods, such as qualitative analysis, quantitative analysis, theoretical analysis, and empirical analysis, to solve the unascertained causal reasoning problem of a reliability evaluation model based on Continuous Bayesian Network.
- 5
- Reliability evaluation of dynamic face recognition systems based on improved Fuzzy Dynamic Bayesian Network5
A novel evaluation method with True Positive Identification Rate in dynamic and M:N mode and create a novel evaluation model of system reliability with the improved Fuzzy Dynamic Bayesian Network to evaluate the reliability of the dynamic face recognition systems, making the evaluation results more reasonable and realistic.
- Cross-Modal Learning Based on Semantic Correlation and Multi-Task Learning for Text-Video Retrieval5
A semantic similarity measurement and semantic consistency classification based on text-video features for a multi-task learning framework that guides the better feature mapping of two modalities and optimizes the construction of unified feature subspace.
- CS-R-FCN: Cross-Supervised Learning for Large-Scale Object Detection5
A novel cross-supervised learning pipeline for large-scale object detection, denoted as CS-R-FCN, to utilize the data flow of image-level annotated images in the fully- supervised two-stage object detection framework, leading to cross- Supervised learning combining bounding-box-levelannotated data and image- level annotated data.
- Tensor index for large scale image retrieval5
A tensor index structure to improve precision and recall simultaneously is proposed and it is shown that on the one hand, the multi-index enhances the discriminative power of visual words, thus improving precision; on the otherhand, the introduction of multiple codebooks corrects quantization artifacts, so improving recall.
- Pose-Free Face Swapping Based on a Deformable 3D Shape Morphable Model5
A pose-free face swapping method based on personalized 3D face modeling based on deformable 3D shape morphable model that outperforms the original color transfer method in dealing with the illumination gap problem is presented.
- 5
- 5
- Beyond Horizontal Paradigm: Circle Anchors for Oriented Object Detection4
This work proposes a novel circle anchor-based representation for oriented object detection, which inherently circumvents regression ambiguity, and introduces circle anchor net, an effective two-stage oriented object detector and develops the circle-to-square dynamic assigner.
- 4
- 4
- 4
- Improving semantic part features for person re-identification with supervised non-local similarity4
A novel part feature learning method, that suppresses the impact of noisy semantic part detection through Supervised Non-local Similarity (SNS) learning, that consistently improves re-ID accuracy using human parsing or pose estimation.
- Person re-identification with effectively designed parts4
This paper benchmarks the accuracy of individual part matching with discriminatively trained Convolutional Neural Network descriptors on the Market-1501 dataset, and investigates the complementarity among different parts using combination and ablation studies, and provides novel insights into this issue.
- 4
- Improved bag-of-words model for person re-identification4
An improved Bag-of-Words (iBoW) model is proposed to describe and match pedestrians under different camera views and is robust against pedestrian variations, and obtained competitive accuracy compared to existing state-of theart methods on person re-identification benchmarks.
- 4
- Person Re-Identification as Image Retrieval Using Bag of Ensemble Colors4
Bag-of-EnsembleColors (BOEC) is proposed to tackle person re-identification, a challenging problem of matching observations of individuals across non-overlapping camera views, and combines low-level color histogram and semantic color names to represent human appearances.
- 4
- Recognition algorithm for sea background infrared targets4
An efficient sub-fitting algorithm was developed to get sea surface information for sea background infrared imagery with low SNR at variant angles and an integrated classifier implements target recognition using the probability-based feature fusion method.
- 4
- 4
- A dynamic generating graphical model for point-sets matching4
A new dynamic generating graphical model for point-sets matching that possesses higher accuracy on point- sets matching and can overcome the limitation of sensitivity on outliers in the existing graphical models is presented.
- 4
- 3
- 3
- 3
- 3
- 3
- 3
- A Digital Twin-based Testing and Data Collection System for Autonomous Driving in Extreme Traffic Scenarios3
A perception-oriented autonomous driving data collection scheme that can store sensor output while running simulation events, and automatically generate corresponding annotations, which can collect data in emergency situations is given.
- Grouping and Decoupling Mechanism for Diabetic Retinopathy Image Grading3
This paper designs a novel deep convolutional neural network to classify DR, which also generates attention map indicating the location of lesions, and outperforms other methods on the Messidor dataset.
- Joint Metric Learning and Hierarchical Network for Gait Recognition3
This paper proposes a novel gait recognition method based on joint metric learning and hierarchical network in an end-to-end manner, for improving the ability of feature identification, and evaluates the proposed method on two widely-used gait datasets, ie.
- 3
- Exploiting EEG Channel Correlations in P300 Speller Paradigm for Brain-Computer Interface3
Experimental results prove the effectiveness and efficiency of the channel correlation analysis based framework and the recognition rate was increased by both 6% with 5 and 10 epochs by the proposed framework.
- Transfer sparse machine: matching joint distribution by subspace learning and classifier transduction3
The proposed TSM consists of two parts: a feature learning part which matches the marginal distributions by learning a shared feature space, and a classifier training part that matches the conditional distributions by training an adaptive classifier in the share feature space.
- A Multi-stage Method for Chinese Text Detection in News Videos3
This paper proposes an improved Stroke Width Transform (SWT) method by incorporating text color consistency constraint for candidate text blocks generation and shows that the proposed method is effective to detect both news video captions and scene texts.
- Realization and Optimization of Embedded Driver Status Detection System3
Experiments on several public databases and driving sequences taken in real car show that the proposed method can detect the driver status precisely and satisfy the real time processing requirements.
- 3
- 3
- 2
- Adaptively Building a Video-language Model for Video Captioning and Retrieval without Massive Video Pretraining2
This paper builds a video-language model in an adaptive manner, which transfers the knowledge from the image domain and can achieve state-of-the-art performance without any further massive video pretraining.
- 2
- Amortized Bayesian Meta-Learning with Accelerated Gradient Descent Steps2
The proposed model aims to solve the training procedure of Bayesian meta-learning to improve the training speed and efficiency, and provides generalization and scalability on unseen samples, and produces competitive/superior uncertainty estimations on few-shot learning tasks on two widely adopted 2D datasets.
- 2
- 2
- 2
- One-Shot Face Recognition Based on Multiple Classifiers Training2
This paper proposes Confidence Constrain Loss to train classifiers in parallel and get better classifiers fusion in the test phase, and uses data augmentation with 3D face reconstruction to obtain a variety of oneshot set's training samples, which can effectively increase the recognition accuracy in the novel set without reducing Recognition accuracy in base set.
- 2
- 2
- 2
- Learning from PhotoShop Operation Videos: The PSOV Dataset2
To demonstrate that the PSOV dataset has sufficient data and labeling for data-driven methods, a deep learning based algorithm for the command classification task is developed.
- Parallel-Structure-based Transfer Learning for Deep NIR-to-VIS Face Recognition2
A Parallel-Structure-based Transfer learning method (PST), which fully utilizes multi-scale feature map information and achieves state-of-the-art performance on CASIA NIR-VIS 2.0 Database, the largest near-infrared face database.
- 2
- 2
- Creating 3D realistic head: from two orthogonal photos to multiview face contents2
The goal of the research is to develop a framework for reconstructing the realistic 3D human head based on two approximate orthogonal views, and the main contribution of the paper is that the processing steps are applies to both the face region as well as the hair region.
- 2
- Multiresolution model construction from scattered range data by hierarchical cube-based segmentation2
- Establishment of a fit prediction model of N95 respirator based on facial images1
A fit prediction model for N95 respirators based on facial images has been developed, which enables the rapid recommendation of N95 respirators with acceptable FF value for healthcare personnel and demonstrates significant potential for practical application.
- 1
- NewsCap: Audio-guided Context-aware Network for News Video Captioning1
A method called Audio-guided Context-aware Network for news video captioning (NewsCap), which consists of two modules named context-aware module which considers comprehensive audio information and audio-guided module which exploits audio information to guide the generation of captions.
- 1
- Controllable and Diverse Image Captioning via Keyword-Driven Diffusion Networks1
It is demonstrated that the capabilities in both diverse and controllable image captioning can be achieved by KCDN with keyword guidance, and it is proposed that the network’s characteristics and proposed hint learning, a strategy that integrates keyword guidance into the training process, can be achieved.
- 1
- 1
- Learning Adversarially Robust Object Detector with Consistency Regularization in Remote Sensing Images1
Experimental results obtained demonstrate that the proposed method effectively improves the robustness of detectors against adversarial attacks.
- 1
- 1
- 1
- 1
- AT-GAN: Attention Transfer GAN for Image-to-Image Translation1
Both qualitative and quantitative experimental results comparison against state-of-the-art methods on benchmark datasets show the superiority of AT-GAN.
- 1
- 1
- A Highly Accurate Feature Fusion Network For Vehicle Detection In Surveillance Scenarios.1
A feature fusion backbone network is proposed to extract vehicle features which has the capability of modeling geometric transformations and a vehicle proposal sub-network is applied to generate candidate vehicle proposals based on multi-level semantic feature maps.
- 1
- Low-Rank Total Variation for Multi-frame Image Super-resolution1
A new algorithm is proposed to restore more high-frequency details by combining both the total variation (TV) and regularization methods, and experimental results show that the proposed algorithm can achieve a better reconstruction quality.
- 1
- 1
- 1
- 1
- Human Action Detection Based on Tracking Region of Maximum Mutual Information1
An algorithm based on tracking of the maximum-mutual-information region of maximum mutual information is presented, which shows the effectiveness of the method in recognizing different actions, and the ability of accurately tracking the action region in large-time-span in the presence of dynamic background.
- 1
- Target recognition in forward looking infrared image based on 3D model1
The experiment result from plenty of real data proves that the proposed method is effective and robust, and performs very well in any case of view point, distance, environment and complex background.
- 1
- A Semi-supervised Clustering-based Segmentation Algorithm of 3D Reconstructed Human Body Parts1
A semi-supervised clustering-based body parts segmentation algorithm of 3D reconstructed human is presented, which can adapt the variety of views, actors and activitiesy and validated with IXMAS database.
- Human Detection with a Coarse-to-fine Cascade Linear SVM1
This work combines the cascade-of-rejection approach with the Histograms of Oriented Gradient (HOG) to form a fast and precise human detector that can process average 12 frames with 320×240 image per second, while maintaining the comparable accuracy to Dalal's method.
- Recognition of multiple 3-D objects based on Markov random field models1
This paper presents an object framework which utilizes densely sampled grids with different resolutions to represent the local information of the input image and shows excellent recognition performance, indicating that the approach is well-suited for appearance-based object recognition.
- The Advance Image Generation Technology and Systems. Scattered Range Data Re-sampling for High-Speed Display of 3-D Objects.1
An efficient method for scattered range data re-sampling by uniform cube-based segmentation that can be reconstructed automatically with fewer patches and displayed in the surroundings of 3-D vision at high speed is proposed.
- –
- –
- –
- Theoretical Framework for the Integration and Development of Digital Economy and Traditional Industries–
Specific policy recommendations were proposed to promote the deep integration and development of China's digital economy and traditional industries from the aspects of infrastructure, data elements, industrial transformation, ecological environment and regional synergy.
- –
- –
- Joint Learning for Scattered Point Cloud Understanding With Hierarchical Self-Distillation–
An end-to-end architecture that compensates for and identifies partial point clouds on the fly and introduces hierarchical self-distillation (HSD), which can be applied to any hierarchy-based point cloud method.
- –
- –
- Enhancing the Encoding Process in Point Cloud Completion–
Experiments show that the proposed hierarchical self-distillation approach outperforms state-of-the-art learning-based methods by at least 1% by incorporating a self-aware loop within the network’s knowledge path.
- Multi-Modality Knowledge Transfer for Generalizable Chest Radiograph Interpretation–
A multi-modal framework that enhances the generalizability of medical imaging models by integrating radiology reports by utilizing a CNN-based visual encoder and transformers to contrast radiographic images and reports is introduced.
- Cosine Similarity Guided Joint Modeling of Foreground and Background for Incremental Object Detection in Drone Images–
This study investigates incremental object detection specifically in aerial scenarios with significant improvements over current state-of-the-art baseline methods under various incremental settings, with the Average Precision improvement by 0.64% and 0.73% in the two-step and multi-step settings, respectively.
- –
- –
- FTAN: Exploring Frame-Text Attention for lightweight Video Captioning–
This paper develops a novel text attention module in FTAN, which uses the hidden state of LSTM as query to generate attentive text features, and is the first to introduce attention mechanism to extract more textual information hidden in LSTM architecture in video captioning.
- –
- Portrait Interpretation and a Benchmark–
The proposed portrait interpretation recognizes the perception of humans from a new systematic perspective, namely Appearance, Posture, and Emotion, and designs corresponding sub-tasks for each aspect.
- –
- –
- –
- –
- Decoupling features via global and local contexts in multi-object tracking–
A feature decoupling module, Global and Local Context-based Decoupling Module (GLCD), to disentangle features extracted by the backbone into two task-specific features, which can choose different contexts by learnable parameters to enforce each self.
- –
- –
- –
- –
- Low-budget real-time pose estimation method for personal physical exercise assistant–
A novel algorithm of pose estimation to determine whether it is correct of doing certain physical exercises by single-person pose-tracking and key frame extraction method is proposed and the proposed tracking compensation method is used to refine the output of Pose estimation network.
- CRRCNN: cascade rotational RCNN for dense arbitrary-oriented object detection–
A novel network named CRRCNN (Cascade Rotational Region-based CNN) to detect dense objects with oriented bounding boxes is presented, which is unified and learned end-to-end by jointly optimizing.
- Hierarchical Structure and Joint Training for Large Scale Semi-supervised Object Detection.–
A novel hierarchical structure and joint training framework for large scale semi-supervised object detection is proposed, utilizing the relationships among target categories to model a hierarchical network to further improve the performance of recognition.
- Exploiting Generalized Matched Filter for Efficient and Effective P300 Detection in Speller Paradigm–
A model based on hypothesis testing is built and a generalized matched filter (GMF) is derived which is a simple linear function of the signal due to the Gaussian noise assumption which improves the performance and enhances the computing efficiency simultaneously.
- –
- On the Large-Scale Transferability of Convolutional Neural Networks–
This paper investigates the effective transfer of the CNN descriptors in generic and fine-grained classification at a large scale and provides some simple yet effective methods in constructing a competitive baseline recognition system.
- –
- –
- An Efficient and Accurate Face Verification Method Based on CNN Cascade Architecture–
An efficient and accurate face verification method based on Convolution Neural Network Cascade architecture that significantly reduces time cost from 485ms to only 20ms on a single core i7-4790, which has strong practical value for real-time face verification systems.
- –
- Metric Learning Based Multi-Patch Ensemble for High Precision Face Verification–
This work proposes an approach that combines a multi-patch deep CNN by using two step metric learning, and shows that this method archives an accuracy of 99.37% on LFW, and 92.80% on YouTube Faces(YTF), which is very competitive with the state of the art.
- –
- –
- Saliency guided region proposal–
This work proposes a novel region proposal framework based on saliency detection to integrate their advantages and shows that this method achieves the superior quality to the state-of-the-arts.
- Discriminative Middle-Level Parts Mining for Object Detection–
This work introduces an unsupervised, exemplar-based training process for part detection, and combines discriminative part mining with the deformable part model (DPM) for object detection.
- –
- New Research Methods for Media and Cognition Experiment Course–
The understanding of the new practical engineering projects on media and cognition course is described, which can improve the students’ ability to grasp related knowledge points and have the ability to plan projects and solve practical problems.
- –
- Combining Fast Extracted Edge Descriptors and Feature Sharing for Rapid Object Detection–
A novel kind of edge/contour descriptors is presented and they serve as the basic features for sharing in feature sharing problem for object detection in cluttered scenes and a mathematical representation of shared features for "sliding-window" based object detection methods is given.
- Pedestrian Detection Based on an Assembly of Sub-image Detectors–
The proposed algorithm has a strong ability to detect the pedestrians, even under a condition of occlusion, and has been tested on different databases and on real scenes.
- –
- –
- Acquisition method of the driver's facial features location–
A method based on mathematical morphology, facial structure and gray integral projection is proposed to achieve the precise positioning of the driver's eyes and the experiment indicates that this method has a high rate of accuracy and robustness.
- Collaborative MR Workspace with Shared 3D Vision Based on Stereo Video Transmission–
This work believes that mixed reality is ideal for collaborative interfaces because it addresses two major issues in three dimensional computer supported collaborative work: cooperation-centred, real object-based operation and network communication.
- Real-time vehicle detection and tracking based on traffic scene analysis–
A new object detection algorithm via fusion of global classifiers and part-based classifier and a vehicle detection algorithm integrating classifying confidence and local shadow is presented, which shows high efficiency and robustness when coping with environment transition, illumination variation and vehicle orientation change.
- Multi-view Moving Human Detection and Correspondence Based on Object Occupancy Random Field–
A novel algorithm is proposed based on object occupancy random field (abbreviated by OORF), which has a robust performance even under severe occlusions and a fast implementation speed.
- AdaBoost tracker embedded in particle filtering–
A novel approach is proposed to fuse the trained classifier into the visual tracker that integrates all weak and strong classifiers at different stages of the trained cascade using AdaBoost into the observation model for particle filtering.
- –
- Object Detection Via Fusion of Global Classifier and Part-Based Classifier–
This paper presents a framework for object detection via fusion of global classifier and part-based classifier, which gives distinctively superior performance than traditional cascade classifiers.
- Recognition of 3-D objects based on Markov random field models–
A new object framework is presented, which utilizes densely sampled grids with different resolutions to represent the local information of the input image and a Markov random field model is then created to model the geometric distribution of the object key nodes.
- Adaptive Sparse Vector Tracking Via Online Bayesian Learning–
A novel real-time tracking method based on online learning with trained RVM classifier embedded into particle filtering for tracking that can efficiently and effectively handle drastic illumination variation, partial occlusion and rapid changes in viewpoint, pose and scale.
- A Framework of Collaborative Workspace Based on Multi-stereo Vision and Shared Mixed Reality–
A framework of collaborative system based on multi-stereo vision for a shared mixed reality (MR) application to show a real world object with a stereo view and natural operating interface for the separated users is proposed.
- –
- Combining local descriptions with geometric constraints for 3D object recognition in multiple statuses–
Comparisons with previous methods show that the proposed Markov random field based framework is far more robust in the presence of object zooming, rotation, occlusion, noise, and viewpoint variations.
- –
- –
- –
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.