Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks423 from public data
- Dual Attention Network for Scene Segmentation6,830
New state-of-the-art segmentation performance on three challenging scene segmentation datasets, i.e., Cityscapes, PASCAL Context and COCO Stuff dataset is achieved without using coarse data.
- 2,068
- 980
- 929
- Unsupervised Feature Selection Using Nonnegative Spectral Analysis545
A new unsupervised learning algorithm, namely Nonnegative Discriminative Feature Selection (NDFS), which exploits the discriminative information and feature correlation simultaneously to select a better feature subset.
- 392
- 371
- 360
- 351
- 316
- CoupleNet: Coupling Global Structure with Local Parts for Object Detection304
This paper proposes a novel fully convolutional network, named as CoupleNet, to couple the global structure with local parts for object detection and achieves state-of-the-art results on all three challenging datasets.
- Hi, magic closet, tell me what to wear!290
This paper collects a large clothing What-to-Wear dataset, and thoroughly annotates the whole dataset with 7 multi-value clothing attributes and 10 occasion categories via Amazon Mechanic Turk, to learn a generalize-well model and comprehensively evaluate it.
- 269
- Decoupled Spatial-Temporal Attention Network for Skeleton-Based Action-Gesture Recognition262
A novel decoupled spatial-temporal attention network (DSTA-Net) for skeleton-based action recognition that involves solely the attention blocks, allowing for modeling spatial-temporal dependencies between joints without the requirement of knowing their positions or mutual connections.
- 256
- Normalized and Geometry-Aware Self-Attention Network for Image Captioning250
This paper proposes Normalized Self-Attention (NSA), a reparameterization of SA that brings the benefits of normalization inside SA and introduces a novel normalization method and demonstrates that it is both possible and beneficial to perform it on the hidden activations inside SA.
- Stacked Deconvolutional Network for Semantic Segmentation223
This work proposes a Stacked Deconvolutional Network (SDN) for semantic segmentation and achieves the new state-ofthe- art results on four datasets, including PASCAL VOC 2012, CamVid, GATECH, COCO Stuff.
- 191
- 176
- 176
- 167
- 167
- Adaptive Context Network for Scene Parsing149
This paper proposes an Adaptive Context Network (ACNet) to capture the pixel-aware contexts by a competitive fusion of global context and local context according to different per-pixel demands.
- 146
- Dual cross-media relevance model for image annotation143
A dual cross-media relevance model (DCMRM) is proposed for automatic image annotation, which estimates the joint probability by the expectation over words in a pre-defined lexicon.
- 142
- 141
- 135
- 127
- 127
- 127
- 117
- 117
- 113
- 112
- 112
- Sparse representation for robust abnormality detection in crowded scenes111
This paper proposes a fully unsupervised non-negative sparse coding based approach for abnormality event detection in crowded scenes, which is specifically tailored to cope with feature noisy and uncertainty.
- 108
- Supervised kernel locality preserving projections for face recognition107
A novel subspace method, named supervised kernel locality preserving projections (SKLPP), is proposed for face recognition, in which geometric relations are preserved according to prior class-label information and complex nonlinear variations of real face images are represented by nonlinear kernel mapping.
- 106
- An adaptive graph model for automatic image annotation105
This paper proposes a novel automatic image annotation method based on manifold ranking learning, in which the visual and textual information are well integrated, and designs a new scheme named the Nearest Spanning Chain (NSC) to generate an adaptive similarity graph.
- Cross-media analysis and reasoning: advances and directions104
By presenting approaches, advances, and future directions in cross-media analysis and reasoning, the goal is to draw more attention to the state-of-the-art advances in the field, and to provide technical insights by discussing the challenges and research directions in these areas.
- Segmentation, categorization, and identification of commercial clips from TV streams using multimodal analysis94
A multimodal ("visual + audio + text") commercial video digest scheme to segment individual commercials and carry out semantic content analysis within a detected commercial segment from TV streams is presented.
- Learning Feature Embeddings for Discriminant Model Based Tracking90
This paper proposes a novel and effective architecture to learn optimal feature embeddings for online discriminative tracking, called DCFST, which integrates the solver of a discriminant model that is differentiable and has a closed-form solution into convolutional neural networks.
- 89
- 89
- 87
- 79
- 78
- Image annotation using multi-correlation probabilistic matrix factorization73
A multi-correlation probabilistic matrix factorization (MPMF) algorithm for the correlation estimation of image-word correlation and the annotation words of each untagged or noisily tagged image can be predicted by reconstructing the image- word correlations with the both derived latent factors.
- 71
- 70
- Non-Autoregressive Image Captioning with Counterfactuals-Critical Multi-Agent Learning69
This paper proposes a Non-Autoregressive Image Captioning (NAIC) model with a novel training paradigm: Counterfactuals-critical Multi-Agent Learning (CMAL), which formulates NAIC as a multi-agent reinforcement learning system where positions in the target sequence are viewed as agents that learn to cooperatively maximize a sentence-level reward.
- Reading scene text with fully convolutional sequence modeling68
An end-to-end fully Convolutional network with the stacked convolutional layers to effectively capture the long-term dependencies among elements of scene text image to achieve state-of-the-art recognition performance and significantly improves the efficiency and reduces the number of parameters.
- AdaSGN: Adapting Joint Number and Model Size for Efficient Skeleton-Based Action Recognition65
A novel approach, called AdaSGN, is proposed in this paper, which can reduce the computational cost of the inference process by adaptively controlling the input number of the joints of the skeleton on-the-fly and can also adaptively select the optimal model size for each sample to achieve a better trade-off between the accuracy and the efficiency.
- 62
- 62
- 61
- Contextual deconvolution network for semantic segmentation59
This paper introduces two types of contextual modules to model the interdependencies of features in channel and spatial dimensions respectively and embeds the two contextual modules into individual components of the decoder network, thus improving the representation power and gaining more precise segment results.
- Community-Based Question Answering via Asymmetric Multi-Faceted Ranking Network Learning59
This paper proposes a novel asymmetric multi-faceted ranking network learning framework for community-based question answering by jointly exploiting the deep semantic relevance between question-answer pairs and the answerers' authority to the given question.
- Multimedia News Summarization in Search58
A novel approach of multimedia news summarization for searching results on the Internet is developed, which uncovers the underlying topics among query-related news information and threads the news events within each topic to generate a query- related brief overview.
- Efficient Clothing Retrieval with Semantic-Preserving Visual Phrases58
Experiments and comparisons indicate that the SPVP significantly enhances the discriminative power of local features with a slight increase of memory usage or runtime consumption compared to the BOW model.
- Coherent bag-of audio words model for efficient large-scale video copy detection57
This paper attempts to tackle the video copy detection task resorting to audio information, which is equivalently important as well as visual information in multimedia processing, and proposes a bag-of audio words (BoA) representation to characterize each audio frame.
- 56
- 55
- Pseudo Label based Unsupervised Deep Discriminative Hashing for Image Retrieval55
Experiments demonstrate that the proposed pseudo label based unsupervised deep discriminative hashing method outperforms the state-of-art unsuper supervised hashing methods.
- 55
- Decoding with Value Networks for Neural Machine Translation54
Experiments show that a prediction network using a recurrent structure for the value network can significantly improve the translation accuracy on several translation tasks.
- Microblog Sentiment Classification via Recurrent Random Walk Network Learning54
This paper proposes a novel recurrent random walk network learning framework for the microblog sentiment classification problem by exploiting both users’ posted tweets and their social relations in microblogs, and introduces the deep recurrent neural networks with random-walk layer for heterogeneous MSC network embedding.
- Towards Unified Multi-Modal Personalization: Large Vision-Language Models for Generative Recommendation and Beyond53
A generic and extensible personalization generative framework, that can handle a wide range of personalized needs including item recommendation, product search, preference prediction, explanation generation, and further user-guided image generation is developed.
- 53
- TopRec52
A unified framework, TopRec, is proposed, which detects topical communities to construct interpretable domains for domain-specific collaborative filtering and a semi-supervised probabilistic topic model is utilized by integrating user guidance with social network.
- 52
- TRAJECT-Bench:A Trajectory-Aware Benchmark for Evaluating Agentic Tool Use51
TRAJECT-Bench is introduced, a trajectory-aware benchmark to comprehensively evaluate LLMs'tool use capability through diverse tasks with fine-grained evaluation metrics, and reveals failure modes such as similar tool confusion and parameter-blind selection, and scaling behavior with tool diversity and trajectory length.
- Structure preserving non-negative matrix factorization for dimensionality reduction51
This paper proposes a novel dimensionality reduction algorithm, called Structure Preserving Non-negative Matrix Factorization (SPNMF), which is formulated as an optimization problem and solved by an effective iterative multiplicative update algorithm.
- 49
- 48
- How friends affect user behaviors? An exploration of social relation analysis for recommendation48
A unified framework to properly incorporate the influence of social relationships into recommendation by the guidance of buddy (friends who have strong influence on user) and susceptibility (the willingness to be influenced) mining is proposed.
- 46
- 46
- A practical SVM-based algorithm for ordinal regression in image retrieval46
This paper focuses on relevance feedback with multilevel relevance judgment, where the relevance feedback is considered as an ordinal regression problem, and proposes a practical SVM-based algorithm for image retrieval upon it.
- Action recognition via pose-based graph convolutional networks with intermediate dense supervision45
A pose-based graph convolutional network (PGCN) is proposed, which encodes the body-part features into a human-based spatiotemporal graph and explicitly models their correlations with a novel light-weight adaptive graph Convolutional module to produce a highly discriminative representation for human action recognition.
- 45
- 45
- MLRank: Multi-correlation Learning to Rank for image annotation43
This paper forms image annotation as a Multi-correlation Learning to Rank (MLRank) problem, i.e., ranking the relevance of tags to an image considering the visual similarity and the semantic relevance, and obtains an explicit solution of the ranking model.
- Automatic composition of broadcast sports video42
This study presents a novel approach to the two major issues required in the system’s implementation, specifically the camera view selection/switching module and the automatic replay generation module, and uses multi-modal framework to perform video content analysis, event and event boundary detection from the raw unedited main/sub-camera captures.
- MC-HOG Correlation Tracking with Saliency Proposal41
A rich feature descriptor, MC-HOG, is proposed, by leveraging rich gradient information across multiple color channels or spaces to effectively explore the color-gradient characteristics and alleviate the model drift problem.
- 41
- 39
- Language Models As Semantic Indexers37
LMIndexer is introduced, a self-supervised framework to learn semantic IDs with a generative language model capable of generating neural sequential discrete representations with progressive training and contrastive learning and is proposed to train with a self-supervised document reconstruction objective.
- 37
- Recommendation by Mining Multiple User Behaviors with Group Sparsity37
A novel recommendation model, the Group-Sparse MatrixFactorization (GSMF), which factorizes the ratingrices for multiple behaviors into the user and itemlatent factor space with group sparsity regularization and can integrate users’ multiple types of behaviors into recommendation better, compared with other state-of-the-arts methods.
- Online video synopsis of structured motion37
An online motion structure preserved synopsis approach is proposed, which can preserve behavior interactions between different objects in the original video while condensing as much content as possible.
- Hierarchical clustering-based navigation of image search results37
HiCluster, an effective method to organize image search results is designed in this paper, which employs both textual and visual analysis and a novel user interface is designed to provide users with the diverse and helpful information based on the hierarchical clustering structure.
- 37
- Highlight ranking for sports video browsing37
This paper presents an advanced highlight analysis system for sports video browsing, in which highlight evaluation and ranking are concerned besides highlight detection, and initial experimental results on soccer videos show an encouraging performance comparing with human subjective evaluation.
- Semantic knowledge extraction and annotation for web images37
By using a page layout analysis method that can precisely assign context to web images, an efficient algorithms to extract semantic knowledge for web image annotation, such as description, people, temporal and geographic information are developed.
- Can Clicks Be Both Labels and Features?36
Experiments on public datasets show that ranking models learned with the proposed framework can significantly outperform models built with raw click features and algorithms that rank items without considering model uncertainty.
- Hierarchically Supervised Deconvolutional Network for Semantic Video Segmentation36
A jointly trained deep learning framework to make the best use of spatial and temporal information for semantic video segmentation and achieves state-of-the-art performance on both of the two datasets.
- Semi-supervised multi-graph hashing for scalable similarity search36
A semi-supervised Multi-Graph Hashing (MGH) framework is proposed that can effectively integrate the multiple modalities with optimized weights in a multi-graph learning scheme and can be more effective for fast similarity search.
- 35
- Tagging Personal Photos with Transfer Deep Learning35
An ontology consisting of a hierarchical vocabulary tailored for personal photos is designed, to solve the semantic distribution gap and discover the intermediate image representations and ontology priors by deep learning with bottom-up and top-down transfers across two domains.
- Boosted multi-class semi-supervised learning for human action recognition35
A boosted multi-class semi-supervised learning algorithm in which the co-EM algorithm is adopted to leverage the information from unlabeled data to deal with the insufficient labeled data and high computational expense is proposed.
- BundleNet: Learning with Noisy Label via Sample Correlations34
BundleNet is proposed, a framework of sequential structure for deep neural networks to handle the label noise, and it is proved that the bundle-module performs a form of regularization, which is similar to dropout as regularization during training, which endows the BundleNet with strong robustness to the label Noise.
- Sketch-based Image Retrieval using Generative Adversarial Networks34
A conditional generative adversarial network is employed to enrich the content information of sketches and recover the imaginary images, and two VGG-based encoders are used to constrain their perceptual consistency from the view of feature representations.
- Improving visual question answering using dropout and enhanced question encoder33
This paper proposes coherent dropout and siamese dropout mechanism to solve the two problems of co-adaptations of neurons and the explosion of output variance in VQA, and develops an enhanced question encoder called Multi-path Stacked Residual RNNs which is deeper and wider and more powerful than current shallow question encoding.
- 33
- A new robust circular Gabor based object matching by using weighted Hausdorff distance33
The experimental results given in this paper show the proposed algorithm is robust to rotation, scale, occlusion, and noise etc.
- Non-Local Graph Convolutional Networks for Skeleton-Based Action Recognition32
A novel two-stream nonlocal graph convolutional network that can be either uniformly or individually learned by BP algorithm, which brings more flexibility and generality for skeleton-based action recognition.
- 32
- Item group based pairwise preference learning for personalized ranking32
A novel PRIGP(Personalized Ranking with Item Group based Pairwise preference learning) algorithm is proposed to integrate item based pairwise preference and item group basedpairwise preference into the same framework.
- 32
- Hashing with dual complementary projection learning for fast image retrieval32
A novel dual complementary hashing (DCH) approach to learn the codes with multiple hash tables so that the binary codes learned by this approach are more powerful for fast similarity search.
- A unified framework for semantic shot representation of sports video32
This paper presents a unified framework for semantic shot representation in field-ball sports genres, in which a video shot is characterized via three essential properties, namely, camera shot size, subject in a scene and video production technology.
- 32
- Texture representation based on pattern map32
A pixel-pattern-based texture feature (PPBTF) which is insensitive to variance of illumination is proposed and compared with multichannel filtering methods show that PPBTF feature is quite time saving and free of the influence of illumination.
- 30
- 30
- Scale multiplication in odd Gabor transform domain for edge detection30
With adjacent scale multiplication in odd Gabor transform domain, a sharpened edge response output is obtained, which can more effectively resist the inverse influence from noise contamination on the performance of edge detector.
- 29
- Real-time people counting for indoor scenes28
A real-time people counting framework for indoor scenes which can count not only moving crowds but also stationary crowds efficiently is proposed and a temporal filter based on static and dynamic crowd analysis is used to further refine the counting results.
- News contextualization with geographic and visual information28
A news browsing system named NewsMap is introduced which can support users in reading news via browsing a map and retrieving news with location queries and enriched images are presented to help users quickly get information.
- 28
- Consumer video retargeting28
A context assisted spatialtemporal grid scheme is presented for consumer video retargeting and a semantic importance map is built up representing the spatial importance and temporal continuity, which is incorporated with a 3D rectilinear grid scaleplate to map frames to the target display.
- A Novel Role-Based Movie Scene Segmentation Method28
The main novelty of this approach is that it converts the movie scene segmentation into a movie-script alignment problem and proposes a HMM alignment algorithm to map the script scene structure to the movie content.
- Automatic TV Logo Detection, Tracking and Removal in Broadcast Video28
A novel and robust framework using tensor based generalized gradient and the OTSU binarization algorithm to logo detection, and a two level framework from coarse to fine to tracking the TV logos.
- 27
- TV ad video categorization with probabilistic latent concept learning27
A bag-of-words representation is proposed to discover ad categories-related latent visual and textual concepts by probabilistic latent semantics analysis (PLSA), which uses multi-modal concepts to represent ad categories in the latent semantics space.
- 27
- DualDS: A dual discriminative rating elicitation framework for cold start recommendation26
A novel Dual Discriminative Selection (DualDS) framework for rating elicitation is proposed, by integrating discriminative selection with dual regularization, and experiments on two real-world datasets demonstrate the effectiveness of DualDS for cold start recommendation.
- Learning Low-Rank Representations with Classwise Block-Diagonal Structure for Robust Face Recognition26
A novel semi-supervised low-rank matrix recovery algorithm for robust face recognition that can learn robust discriminative representations for both training images and testing images simultaneously by exploiting the classwise block-diagonal structure.
- Enhancing news organization for convenient retrieval and browsing26
A news search and browsing system named GeoVisNews is designed, in which the news elements of “Where”, “Who’s’, ‘What’ and “When’ are enhanced via news geo-localization, image enrichment and joint ranking, respectively, to facilitate users to access news quickly and comprehensively.
- A generic framework for event detection in various video domains26
A generic event detection approach based on semi-supervised learning and Internet vision is presented and a novel Multiple Instance Learning Induced Similarity (MILIS) measure by learning instance sensitive classifiers is proposed.
- 25
- Dynamic scene understanding by improved sparse topical coding25
A novel unsupervised approach to automatically explore motion patterns occurring in dynamic scenes under an improved sparse topical coding (STC) framework that is suitable for very large range applications of scenarios, such as rule mining and abnormal event detection.
- Scale-Adaptive Deconvolutional Regression Network for Pedestrian Detection24
This paper proposes a novel Scale-Adaptive Deconvolutional Regression (SADR) network, which can effectively detect pedestrians of various scales by flexibly choosing which feature layer to regress object locations according to the height of pedestrians, thus improving the detection accuracy significantly.
- Semi-supervised Unified Latent Factor learning with multi-view data24
This paper presents a Semi-supervised Unified Latent Factor learning approach (SULF) to learn a predictive unified latent representation by leveraging both complementary information among multiple views and the supervision from the partially label information.
- 24
- Erasing-based Attention Learning for Visual Question Answering23
This paper proposes to enforce the correlation between the attention and the non-attention parts as a constraint for attention learning, and adopts an attention-guided erasing scheme and associates a typical classification loss with the above distance constraint to learn a more discriminative attention map for answer prediction.
- Deep People Counting with Faster R-CNN and Correlation Tracking23
This paper proposes a method to people counting based on Faster R-CNN head-shoulder detection and correlation tracking, which fuse the two bounding box to obtain a continuous and stable trajectory.
- Learning a Representative and Discriminative Part Model with Deep Convolutional Features for Scene Recognition23
A unified framework for learning a representative and discriminative part model with deep convolutional features for scene recognition task and achieves state-of-the-art performances on the standard scene benchmarks.
- 23
- 23
- Image enhancement for outdoor long‐range surveillance using IQ‐learning multiscale Retinex21
A blind image quality learning-based multiscale Retinex is proposed, which will compute its IQ metrics and estimate the MOCPs for the WT_MSR by BPNN, and a kind of optimal enhancement can be realised.
- Objectness-aware Semantic Segmentation21
An objectness-aware semantic segmentation framework (OA-Seg) is proposed by jointly learning an object proposal network (OPN) and a lightweight deconvolutional neural network (Light-DCNN) to provide a finer upsampling way than FCN.
- Multiple deep features learning for object retrieval in surveillance videos21
This study presents an effective multiple deep features learning approach for object retrieval in surveillance videos based on the discriminative convolutional neural network (CNN), which can learnmultiple deep features to comprehensively describe the visual object.
- 21
- Learning Multi-view Deep Features for Small Object Retrieval in Surveillance Scenarios21
This paper proposes to leverage discriminative Convolutional Neural Network to learn deep structure and color feature to form an efficient multi-view object representation and utilizes CNN trained on ImageNet to abstract rich semantic structure information.
- 21
- Node-wise Filtering in Graph Neural Networks: A Mixture of Experts Approach20
This work theoretically demonstrate that a global filter optimized for one pattern can adversely affect performance on nodes with differing patterns, and introduces a novel GNN framework Node-MoE that utilizes a mixture of experts to adaptively select the appropriate filters for different nodes.
- Incremental Matrix Factorization via Feature Space Re-learning for Recommender System20
Using the prediction error of the new data as the auxiliary features, the method updates both feature matrices simultaneously, and thus users' preference can be better modeled than merely adjusting one corresponded feature matrix.
- Group latent factor model for recommendation with multiple user behaviors20
A novel recommendation model, named Group Latent Factor Model (GLFM), which attempts to learn a factorization of latent factor space into subspaces that are shared across multiple behaviors and subsp spaces that are specific to each type of behaviors, which can integrate users' multiple types of behaviors into recommendation better.
- 20
- Cross-media retrieval: state-of-the-art and open issues20
According to the current technologies used in cross-media retrieval and the demand from real-world applications, open research issues and future research opportunities are identified.
- 19
- 19
- Object co-segmentation via salient and common regions discovery19
An adaptive discriminative low rank matrix recovery (ADLRR) algorithm to divide the over-completely segmented regions of a given image set into object and non-object ones and an efficient alternate optimization procedure based on block-coordinate descent and augmented Lagrange multipliers method is designed.
- Learning Binary Codes with Bagging PCA19
This paper attempts to leverage the bootstrap sampling idea and integrate it with PCA, resulting in a new projection method called BaggingPCA, in order to learn effective binary codes, closely connected with the core idea of LSH.
- Low rank metric learning for social image retrieval19
A Low Rank distance Metric Learning (LRML) algorithm by discovering knowledge from these rich contextual data, to boost the performance of CBIR and formulate it as a convex optimization problem with nuclear norm minimization and propose an effective optimization algorithm based on proximal gradient method.
- 18
- Social recommendation via multi-view user preference learning18
This paper introduces the social recommendation (SR) problem utilizing microblogging data and develops an efficient alternating direction method of multipliers (ADMM) scheme to solve the problem via multi-view user preference learning.
- Key observation selection-based effective video synopsis for camera network18
A novel video synopsis algorithm for partially overlapping camera network that can generate video synopsis for large scenarios, which can facilitate understanding overall activities and adopt a novel unsupervised graph matching algorithm to associate trajectories across cameras.
- 18
- Personalized retrieval of sports video18
A novel approach is proposed to achieve personalized retrieval of sport video, which includes two research tasks: semantic annotation of sports video and acquisition of user's preference, which has an encouraging performance.
- 18
- A unified model sharing framework for moving object detection17
This paper proposes a unified model sharing framework for moving object detection that reduces the number of models and enhances precision, and embeds existing approaches and improves their performance.
- Correlation consistency constrained probabilistic matrix factorization for social tag refinement17
A social tag refinement method, named as Correlation Consistency constrained Probabilistic Matrix Factorization (CCPMF), to jointly model the inter- and intra-correlations among images and tags, and further to precisely reconstruct the image-tag correlation as a result.
- Size Adaptive Selection of Most Informative Features17
A novel method to select the most informatives subset of features, which has little redundancy and very strong discriminating power is proposed, and the obtained classification results significantly outperform the state-of-the-art results on most test datasets.
- 17
- 17
- 17
- 17
- Deep learning driven hypergraph representation for image-based emotion recognition16
This paper proposed a bi-stage framework for image-based emotion recognition by combining the advantages of deep convolutional neural networks (D-CNN) and hypergraphs and tackled the emotion prediction problem by a transductive learning approach.
- Chat with illustration16
A novel visual assisted instant messaging scheme named Chat with illustration (CWI), which presents users visual messages associated with textual message automatically, and a visual dialogue summarization is also proposed, which help users recall the past dialogue.
- Human behaviour consistent relevance feedback model for image retrieval16
This paper proposes a human behavior consistent relevance feedback model for image retrieval that adopts the graph ranking algorithm to model the retrieval process and demonstrates the effectiveness of the proposed mechanism.
- 16
- Kernel-based nonlinear discriminant analysis for face recognition16
A nonlinear subspace analysis method, Kernel-based Nonlinear Discriminant Analysis (KNDA), is presented for face recognition, which combines the nonlinear kernel trick with the linear sub space analysis method — Fisher Linear Discriminatory Analysis (FLDA).
- 16
- Reading selectively via Binary Input Gated Recurrent Unit15
Inspired by human reading, binary input gated recurrent unit (BIGRU), a GRU based model using a binary input gate instead of the reset gate in GRU is introduced, which can read selectively during interference and achieves better performances than baseline GRU in all the testing tasks.
- Interactive ads recommendation with contextual search on product topic space15
A ad domain based concept hierarchy is built to make the most of the product details over the e-commerce sites to provide effective advertising suggestions and users can interactively select keyframes or keywords to personalize their intentions by textual re-search.
- Event tactic analysis based on player and ball trajectory in broadcast video15
This paper proposes a novel approach to extract the tactic information and recognize the tactic patterns from goal events in broadcast soccer video using the analysis and alignment of web-casting text and broadcast video.
- 15
- 15
- A non-parameter bayesian classifier for face recognition15
A non-parameter Bayesian classifier based on Kernel Density Estimation (KDE) is presented for face recognition, which can be regarded as a weighted Nearest Neighbor (NN) classifier in formation.
- 14
- 14
- 14
- Nonlinear matrix factorization with unified embedding for social tag relevance learning14
A nonlinear matrix factorization method with the priors of inter- and intra-correlations among images and tags to effectively predict the tag relevance to the visual contents is proposed.
- Effective logo retrieval with adaptive local feature selection14
A novel approach to extract and combine local features for effective logo retrieval by developing a strategy of adaptive feature selection by a weight updating mechanism is proposed.
- Improvement of Shot Detection Using Illumination Invariant Metric and Dynamic Threshold Selection14
color ratio histograms are used as frame content measure and a local adaptive threshold technique is adopted to utilize the local characteristic of shot change, and the effectiveness of the methods is validated by experiments on some real-world video sequences.
- 3D Object Retrieval with Multimodal Views13
The evaluation results show a promising scenario about multimodal view-based 3D retrieval methods, and reveal interesting insights in dealing with multi-modal data.
- 13
- 13
- 13
- 13
- Personalized retrieval of sports video based on multi-modal analysis and user preference acquisition13
A novel framework on personalized retrieval of sports video, which includes two research tasks: semantic annotation and user preference acquisition, and an encouraging performance of the proposed method is shown.
- Visual Tracking Using Particle Filters with Gaussian Process Regression13
A novel visual-tracking algorithm using particle filters with Gaussian process regression and resampling techniques, which effectively abate the influence of particle degeneracy and improve the robustness of visual tracking.
- Automatic Watermeter Digit Recognition on Mobile Devices12
This paper adopts Adaboost with aggregated channel features (ACF) to detect watermeter digital regions, and designs a small attention bidirectional long short-term memory (BLSTM) for end-to-end digit sequence recognition.
- Supervised Hashing with Soft Constraints12
A general framework for supervised hashing is presented which does not toughly require a dissimilar pair to have maximum Hamming distance, and a soft constraint which can be viewed as a regularization to avoid over-fitting is utilized.
- Ordinal regularized manifold feature extraction for image ranking12
A novel feature extraction algorithm specifically designed for learning to rank in image ranking not only targets at preserving the local manifold structure of data, but also keeps the ordinal information among different data blocks in the low-dimensional subspace, where a ranking model can be learned effectively and efficiently.
- Anomaly detection in crowded scene via appearance and dynamics joint modeling12
A novel high-frequency feature based on optical flow (HFOF) is introduced and multiple kernel learning (MKL) is adopted to train a classifier for anomaly detection.
- 12
- 12
- 12
- Progressive rectification network for irregular text recognition11
This paper proposes a novel progressive rectification network (PRN) for irregular scene text recognition that progressively rectifies the irregular text to a front-horizontal view and further boosts the recognition performance.
- 11
- ActiveAd: A novel framework of linking ad videos to online products11
This paper is the first attempt to improve the effectiveness of advertising through combining online shopping information with an ad video and directing viewers to proper online shopping places and achieves product recommendation effectively and efficiently.
- A Real-Time People Counting Approach in Indoor Environment11
A block-updating way to update the background model, a multi-view head-shoulder model to find candidate people, and an improved k-means clustering is proposed to locate the position of each people are designed.
- Projective Matrix Factorization with unified embedding for social image tagging11
This paper presents a general formulation, named Projective Matrix Factorization with unified embedding (PJMF), by which social image retagging is transformed to the nearest tag-neighbor search for each image.
- Asymmetric propagation based batch mode active learning for image retrieval11
A novel batch mode active learning scheme for informative sample selection inspired by the method of graph propagation is presented, which takes the correlation between labeled samples and unlabeled samples, but the correlation among unlabeling samples taken into account as well.
- Correlated PLSA for Image Clustering11
An improved PLSA model, named Correlated Probabilistic Latent Semantic Analysis (C-PLSA), where the topics of the given image are modeled by the images that are related to it, and the approach is superior to the traditional PLSA for image clustering.
- 11
- 11
- 11
- Providing on-demand sports video to mobile devices11
An infrastructure for extracting and delivering the highlights instead of the whole sport videos to mobile clients, which can significantly reduce the bandwidth consumption is constructed and an advanced UI for the mobile clients to effectively browse and interact with the video highlights is designed.
- Personalized Recommendation Meets Your Next Favorite10
This paper proposes a unified model, namely States Transition pAir-wise Ranking Model (STAR), to address users' favorites mining for sequential-set recommendation, and develops a pair-wise ranking loss function for parameters learning.
- A new extension of kernel feature and its application for visual recognition10
The kernel feature vector of an image is extended to a kernel feature matrix for visual recognition using multiple similarities to measure two images, which is more accordant to human vision.
- 10
- Selective Sampling Based on Dynamic Certainty Propagation for Image Retrieval10
A novel Dynamic Certainty Propagation (DCP) scheme for informative data selection that only labels one single data point with the lowest degree of certainty each time, and updates the rest unlabeled data's certainty dynamically according to their correlation.
- 10
- 10
- 10
- 10
- Multilevel Relevance Judgment, Loss Function, and Performance Measure in Image Retrieval10
A new relevance feedback scheme based on a support vector learning algorithm for ordinal regression and a performance measure that is based on the preference of one image to another one are developed.
- Random subspace for binary codes learning in large scale image retrieval9
This work introduces a random subspace strategy to address the limitation of hashing based approximate nearest neighbor search methods, which leads to an unexpected phenomenon that longer hashing code does not necessarily yield better performance.
- Sparse semantic metric learning for image retrieval9
A sparse semantic metric learning (SSML) algorithm is proposed by discovering knowledge from social media resources, and applying the learned metric to search relevant images for users by exploiting heterogeneous information from two views of images and formulates the learning problem with the following principles.
- 9
- 9
- 9
- 9
- 9
- Community-Based Question Answering via Contextual Ranking Metric Network Learning8
This paper proposes a novel framework for CQA task by exploiting both the question-answer content in C QA site and users' social contexts, and shows the effectiveness of this method on real-world dataset.
- Semi- and Weakly- Supervised Semantic Segmentation with Deep Convolutional Neural Networks8
A semi- and weakly-supervised learning framework by exploring images most only with image-level labels and very few with pixel- level labels, in which two stages of Convolutional Neural Network (CNN) training are included.
- Automatic face annotation in TV series by video/script alignment8
This paper considers the temporal order relationship of the faces and names over the whole video and directly align two heterogeneous sequences to annotate the faces by video/script alignment.
- Object co-segmentation via discriminative low rank matrix recovery8
A Discriminative Low Rank matrix Recovery (DLRR) algorithm to divide the over-completely segmented regions of a given image set into object and non-object ones is proposed and an efficient optimization procedure based on block-coordinate descent is designed.
- A Weighted One Class Collaborative Filtering with Content Topic Features8
This paper proposes to improve recommendation by considering the rich user and item content information to assist weighting the unknown data in OCCF, and gets a content topic feature for each users and item by using probabilistic topic modeling method.
- 8
- Boosting part-sense multi-feature learners toward effective object detection8
A novel comprehensive learner design mechanism toward effective object detection in terms of both discrimination and generalization abilities is proposed, which outperforms the traditional single-feature learner and other multi-feature learners in discriminative andgeneralization abilities.
- 8
- 8
- A Fast Algorithm for Delaunay Based Surface Reconstruction8
A nonuniformly sampling method is proposed to resample the input data set according to the local feature size before reconstruction to increase the speed of reconstruction without losing the details the authors need.
- Mosaic representations of video sequences based on slice image analysis8
The proposed method is based on slice image analysis, and can effectively avoid the local minimum problem and remove the accumulated distortions from the final mosaic which are common in the previous methods.
- 7
- On Lemon Defect Recognition with Visual Feature Extraction and Transfers Learning7
A convolutional neural network with an embedded Visual Geome-try Group 16 based (VGG16-based) network using transfer learning is constructed to recognize defect lemons.
- Progressive Bi-C3D Pose Grammar for Human Pose Estimation7
A progressive pose grammar network learned with Bi-C3D (Bidirectional Convolutional 3D) for human pose estimation achieves competitive performance on both MPII and LSP benchmarks compared with previous methods, which confirms the feasibility and effectiveness of C3D in information interactions.
- 7
- Learning representative and discriminative image representation by deep appearance and spatial coding7
A generalized BoF-based framework, which is hierarchically learned by exploring recently developed deep learning methods, and a structured sparse Auto-encoder is proposed to explore the region representations into the image-level signature.
- 7
- 7
- M4L: Maximum margin Multi-instance Multi-cluster Learning for scene modeling7
M^4L is further relaxed and solved by making use of a combination of the Cutting Plane method and Constrained Concave-Convex Procedure, and the results validate the effectiveness of the proposed approach.
- 7
- Visual Attention Model Based Object Tracking7
A biological visual attention based object tracking algorithm that combines the top-down, task dependent attention and bottom-up, stimulus driven attention and can be integrated into other object tracking methods as a prior to increase the robustness and efficiency of tracking.
- Fast feature selection and training for AdaBoost-based concept detection with large scale datasets7
Experimental results reveal the method can significantly reduce the training time of the best learner searching procedure, and the performance of the method is comparable with the state-of-art methods.
- Building topographic subspace model with transfer learning for sparse representation7
A topographic subspace learning algorithm, named key-coding learning, which utilizes irrelevant unlabeled auxiliary data to facilitate image classification and retrieval tasks, and generates the succinct sparse representations by mapping target data into this topographic model.
- 7
- Particle-filter-based object tracking with color and texture information fusion7
Experimental results demonstrate that the performance of the proposed method is more robust and accurate than the original color based method, especially when tracking objects with similar color appearance to the background and partial occlusions.
- 7
- 7
- 7
- 7
- 7
- Consistent-Separable Feature Representation for Semantic Segmentation6
This paper develops a Consistent-Separable Feature Learning Module to obtain C-S features through a new loss, called Class-Aware Consistency loss, proposed to force the deep features to be consistent among the same class and apart between different classes.
- 6
- Clustering based ensemble correlation tracking6
A clustering based ensemble correlation tracker is proposed to jointly capture the target appearance by multi-scale kernelized correlation filter and exploit the long-term object properties by the object template with cluster analysis.
- Learning weighted part models for object tracking6
This paper presents a weighted part model tracker that can efficiently handle partial occlusion and appearance change and presents a temporal weighted model to dynamically adjust the importance of each part by measuring the fitness to the historical temporal distributions in the tracking process.
- What Visual Attributes Characterize an Object Class?6
The approach ensures the learned visual attributes to be visually recognizable and representative, in contrast to manually constructed attributes that contain properties difficult to be visualized, e.g., “smelly,” “smart.”
- 6
- Adaptive spatial partition learning for image classification6
This work proposes a data-driven approach to adaptively learn the discriminative spatial partitions corresponding to each class, and explore them for image classification, and adopts a discrim inative learning formulation with the group sparse constraint to find a sparse mapping from the feature representation to the label space.
- Personalized Sports Video Customization Using Content and Context Analysis6
A user-participant multiconstraint 0/1 Knapsack model is designed to model the personalized video customization, which can unify both video retrieval and summarization with different fusion parameters.
- 6
- 6
- 6
- 5
- 5
- Enhancing Visual Question Answering Using Dropout5
It is shown that the coherent dropout is simple in implementation but very effective to overcome overfitting and the siamese dropout mechanism can bring additional improvements on the state-of-the-art VQA models.
- Image Classification Using Spatial Difference Descriptor Under Spatial Pyramid Matching Framework5
This work exploits a novel descriptor called spatial difference to concatenate the histograms of bag-of-visual words model under spatial pyramid matching framework and conducts image classification experiments on several public datasets to demonstrate the effectiveness of the proposed scheme.
- A hybrid domain enhanced framework for video retargeting with spatial–temporal importance and 3D grid optimization5
A hybrid framework of video retargeting with a domain enhanced spatial-temporal grid optimization and a semantic importance map is built up representing the spatial importance and temporal continuity, which is incorporated with a 3D rectilinear grid scaleplate to map frames to a target display.
- Multiple features fusion for crowd density estimation5
A crowd density estimation approach fusing statistic features and texture analysis was proposed and after extracting foreground objects with frame difference, SVM classifiers with GLCM and statistical features are learned.
- Snap & play5
New mobile game design principles and techniques for enhancing player's gaming experience in personalized, automatic, and dynamic aspects are explored by taking the popular game, Find-the-Difference (FiDi), as a concrete example.
- Adaptive Model for Robust Pedestrian Counting5
A model-based approach based on a whole pedestrian model built up in advance, adaptive models are dynamically determined by the occlusion conditions of corresponding body parts to examine visibility of certain body part.
- Personalized Sports Video Customization for Mobile Devices5
The proposed mobile personalized sports video customization system aims to provide mobile users with interesting video clips according to their personalized preferences and is proved effective by both quantitative and qualitative experiments conducted on five sports matches.
- A Hierarchical Semantics-Matching Approach for Sports Video Annotation5
A hierarchical semantics-matching approach to annotate sports video to link video and text with high-level semantics rather than low-level features and find the optimal video-text alignment based on the integral structure rather than individual conditions.
- 5
- 5
- Image Segmentation Using Co-EM Strategy5
An image segmentation algorithm using co-EM strategy, which employs two views instead of conventional single view based EM to estimate the parameters of GMM, and outperforms the single viewbased GMM-EM segmentation methods.
- Unsupervised change detection on SAR images using Markovian fusion5
The fusion approach based on MRFs model is a promising way of achieving robust unsupervised change detection in temporal sets of synthetic aperture radar (SAR) images using Markovian fusion.
- Motion Detection in Driving Environment Using U-V-Disparity5
An efficient solution for motion detection in driving environment using a stereovision based method based on U-V-disparity concept, which has been tested on real road sequences and experimental results verified its efficiency.
- Fusion Method of Fingerprint Quality Evaluation: From the Local Gabor Feature to the Global Spatial-Frequency Structures5
A new fusion method to evaluate fingerprint quality by combining both spatial and frequency features of a fingerprint image, which provides a feasible rejection of poor fingerprint images before they are presented to the fingerprint recognition system for feature extraction and matching.
- A collaborative environment for enhanced information access on small-form-factor devices5
A distributed user interface that crosses devices to cooperatively present information adaptable to small displays is designed and applied to the document and webpage that are ubiquitously available media content on mobile devices.
- 5
- MENet: A Memory-Based Network with Dual-Branch for Efficient Event Stream Processing4
A novel memory-based network with dual-branch structure, namely MENet, that significantly reduces the computational complexity compared with previous methods while achieving state-of-the-art performance on gesture recognition and object recognition.
- BTDP4
It is proved that using such a block-diagonal core tensor is equivalent to conducting many “tiny” bilinear operations in different feature spaces, which can significantly increase the performance of feature fusion and improve VQA models.
- Learning discriminative context models for concurrent collective activity recognition4
A unified discriminative learning framework of multiple context models is proposed for concurrent collective activity recognition and jointly model the multiple context cues (intra-class, inter-class and global-context) with a max-margin leaning framework.
- 4
- Part Context Learning for Visual Tracking4
A unified context learning framework that can capture stable structure relations of in-object parts, context parts and the object itself to enhance the tracker’s performance is proposed.
- Mask Assisted Object Coding with Deep Learning for Object Retrieval in Surveillance Videos4
A novel approach to efficiently encode and retrieve visual objects, which addresses some practical complications in surveillance videos, is presented, which significantly outperforms state-of-the-art solutions for object retrieval in large-scale video dataset.
- Finding logos in real-world images with point-context representation-based region search4
This paper proposes a novel framework for finding visual logos in the real-world images by exploiting the contextual shape and patch information around feature points, merge them into a combined feature representation (point-context).
- Point-context descriptor based region search for logo recognition4
A point-context descriptor for region modeling is adopted, which is a highly correlated integration of three kinds of features: point, shape, and patch, for logo recognition.
- Dual local consistency hashing with discriminative projections selection4
This paper proposes a dual local consistency hashing method which not only makes the similar images have the same codes but also dissimilar images with different codes, and proposes a PCA projection selecting scheme that choose the most discriminative projection for each bit of the codes.
- Specific vehicle detection and tracking in road environment4
A novel background subtraction approach is used to get the accurate foreground segmentations and shadow suppression and a HIK (Histogram Intersection Kernel) based SVM classifier is trained to recognize whether a vehicle is suspicious.
- Feature selection under learning to rank model for multimedia retrieve4
This paper utilized machine learning techniques to automatically select useful features for ranking by generating a set of feature subsets and sorting them according to the ranking performances.
- 4
- 4
- 4
- COMMERCIAL VIDEO RETRIEVAL WITH VIDEO-BASED BAG OF WORDS4
A commercial video with video-based bag of words mode and FMPI concepts is presented, and a coarse-to-fine framework for robust commercial retrieval in video streams is proposed, showing the promise of the proposed algorithm.
- A Semantic Image Category for Structuring TV Broadcast Video Streams4
This paper proposes a semantic image category, named as Program Oriented Informative Images (POIM), to facilitate the segmentation, indexing and retrieval of different programs in TV broadcast video stream.
- 4
- 4
- 4
- Random Independent Subspace for Face Recognition4
A novel method is proposed to enhance ICA classifier for the small sample size problem by using the random resampling method to generate some random independent subspaces, and a classifier is constructed in each subspace.
- 4
- Asynchronous Event Processing with Local-Shift Graph Convolutional Network3
A local-shift graph convolutional network (LSNet) is proposed, which utilizes a novel local- shift operation equipped with a local spatio-temporal attention component to achieve efficient and adaptive aggregation of neighbor features to improve the efficiency of pooling operation in feature extraction.
- Piecewise Video Condensation for Complex Scenes3
This paper introduces a feedback scheme and a selective diffusion strategy to keep the integrity of foreground objects, followed by a sticky trajectory method to remove noisy fragments and reduce blinking effect, and an adaptive truncation strategy to raise the condensation ratio and improve the visual quality.
- 3
- 3
- Exclusive Constrained Discriminative Learning for Weakly-Supervised Semantic Segmentation3
This paper designs an effective but simple weakly-supervised constraint, and proposes an exclusive constrained discriminative learning model for image semantic segmentation that can be easily optimized via alternative iterations.
- 3
- 3
- Real-Time Cascade Template Matching for Object Instance Detection3
Traditional object instance detection methods are mostly based on template matching, which means that they find where a specific object instance is in an image or a video frame.
- Real-time multiple object instances detection3
A novel, real-time multiple object instance detection system via template matching and pairwise classification that makes use of a pairwise-based classifier across instance categories to test and verify these candidates with respect to templates.
- 3
- People Detection by Boosting Features in Nonlinear Subspace3
A novel approach to detect people by boosting features in the nonlinear subspace by combining three types of the HOG descriptor, which is extracted and grouped into one descriptor to represent the samples.
- A Survey on Sports Video Analysis3
This paper gives a review of relative research work including low-level feature extraction, mid-level keyword generation, high-level semantics inference, relative applications and system prototypes, and finally indicates potential trend.
- IVA-NLPR-IA-CAS TRECVID 2009: High Level Features Extraction3
An overview and comparative analysis of the HLF detection system is presented and average fusion of 4 SVM classification results for each concept using various feature representation choices are presented.
- 3
- 3
- 3
- Robust Speaking Face Identification for Video Analysis3
It is demonstrated that high precision can be achieved by aligning mouths through face matching, which needs no accurate mouth location.
- Hybrid approach for remote sensing image registration3
A hybrid registration method is proposed that uses the registration approach based on SIFT to get the initial parameters, and then area-based method is employed to refine the performance of registration to improve the efficiency of computation.
- Multiple Similarities Based Kernel Subspace Learning for Image Classification3
A revised version of Two-Dimensional PCA algorithm is developed to learn intrinsic subspace of matrix features for classification, in which matrix based kernel features are designed to capture the multiple similarities between images in different low-level visual cues.
- 3
- Image Retrieval Based on Independent Components of Color Histograms3
ICA is applied to reduce the dimensionality and then ICA is performed on the low-dimensional PCA subspace and results show that the proposed method outperform other methods based on SVD of quadratic matrix or PCA, in terms of retrieval accuracy.
- 2
- 2
- Enhanced Bounding Box Estimation with Distribution Calibration for Visual Tracking2
DCOM is presented, a novel bounding box estimation method for visual tracking, based on distribution calibration and overlap maximization, that can be built on top of off-the-shelf networks without finetuning and extra parameters.
- Gate-based Bidirectional Interactive Decoding Network for Scene Text Recognition2
A novel Gate-based Bidirectional Interactive Decoding Network for scene text recognition that combines a gated context mechanism to adaptively make use of the visual context and reverse language context and relieve the adverse effect of noises.
- Heterogenous Graph Mining for Measuring the Impact of Research Institutions2
The proposed scheme involved in the KDD Cup 2016 mainly comprises of feature engineering and application of decision tree models and resulted in the first place in phase 2 and fourth place in overall rank.
- 2
- Object Categorization Using Local Feature Context2
This paper proposes a novel object categorization method by considering the local feature context, which uses theLocal feature context to represent this position’s visual information and is also invariant to rotation and scale change.
- Modeling Hidden Topics with Dual Local Consistency for Image Analysis2
A novel topic model is presented, which can learn an effective and robust mid-level representation in the latent semantic space for image analysis and understanding and considers both the local image structure and local word consistency simultaneously when estimating the probabilistic topic distributions.
- 2
- Ordinal preserving projection2
A novel dimensionality reduction algorithm named ordinal preserving projection (OPP) for learning to rank, which aims at leveraging the global structure of the data set and ordinal information of the observations to obtain two optimal projection matrices.
- 2
- Audio-visual large-scale video copy detection2
A bag of visual concept words model for video copy detection method is proposed from visual cue, and the main advantage of the proposed method lies in the fact that the presence of semantic concepts is highly robust to spatial and temporal video transformations.
- Linking Traditional TV Advertising to Internet Advertising2
This work presents the current research efforts in the field of analyzing TV ads as well as linking TV ads with relevant Internet ads, and suggests how to collect correlated advertisements across multiple media platforms such as TV and Internet would see significant growth potential in cross-media marketing campaigns.
- Human action recognition via multi-view learning2
This paper is the first to try the Co-EM based multi-view learning method for action recognition and obtain significantly better results, and extensively verify the proposed approach on two publicly available challenging datasets.
- Human action recognition in videos using motion impression image2
A new method to recognize human action using motion information in video by combining two types of impression images from different views into a spatial pyramid matching kernel (SPMK) based classifier to recognize various human actions.
- Discovering tactics in broadcast sports video with trajectories2
A generic scheme to discover tactics in classified trajectories in the same events using frequent pattern mining and Transductive classifier with the learnt bias of low-level features of the trajectories is introduced.
- 2
- Concept-Specific Visual Vocabulary Construction for Object Categorization2
This work proposes a novel method to construct a concept-specific visual vocabulary by extracting various visual features from local image patches, and cluster them separately according to different features to generate an initial vocabulary.
- 2
- 2
- A Cooperative Framework for Information Browsing in Mobile Environment2
A cooperative framework to facilitate information browsing among devices in mobile environment is proposed by establishing a communication mechanism to maintain cooperative browsing across devices and designing a distributed user interface across devices to overcome the small display area limited by a single device.
- Boosting relative spaces for categorizing objects with large intra-class variation2
A novel method for object categorization is proposed, where the weighted FLDs as weak learners trained in relative spaces are integrated with the boosting framework to form the final classifier.
- 2
- 2
- Phase-only transform based shaping for error concealment coding2
A novel approach of error resilient coding is proposed by introducing a phase-only transform to the low bit MDVQ (Multiple description vector quantization) method, which outperforms several other approaches such as pixel interleaving,MDVQ and phase scrambling in PSNR.
- 2
- Head tracking using shapes and adaptive color histograms2
The method is demonstrated to be a real-time tracker and robust to clutter, scale variation, occlusion, rotation and camera motion, for several test sequences.
- An ICA-Based Illumination-Free Texture Model and Its Application to Image Retrieval2
A novel pixel pattern-based approach for texture classification, which is independent of the variance of illumination, is proposed, and Comparisons with multichannel and multiresolution features indicate that the new feature is quite time saving, free of the influence of illuminate, and has notable accuracy.
- 1
- SC4D: A Sparse 4D Convolutional Network for Skeleton-Based Action Recognition1
A novel sparse 4D convolutional network (SC4D) is proposed to directly process the generated 4D grid for high-level perceptions and makes better use of the advantages of the convolved network, resulting in a more concise, general and robust framework for skeletal data.
- 1
- Answer Distillation for Visual Question Answering1
A method named answer distillation is proposed to decrease the scale of answer space and limit the correct result into a small set of answer candidates, providing a new state-of-the-art performance on COCO-VQA dataset.
- Sparse Overlap Cross-Platform Recommendation Via Adaptive Similarity Structure Regularization1
A cross-platform recommendation framework termed Adaptive Similarity Structure Regularization Through Connector (AdaSTC), which adaptively learns the user similarity structure on different platforms and further uses it to regularize the modeling process of user preference.
- Insect Recognition Using Sparse Coding and Decision Fusion1
Experimental results on a Tephritidae dataset show that the three proposed discriminative coding methods perform well in insect recognition, and the proposed fusion scheme improves the recognition accuracy significantly.
- 1
- Collaborative Tracking: Dynamically Fusing Short-Term Trackers and Long-Term Detector1
This paper proposes a novel collaborative tracking framework fusing short-term trackers and long-term object detector, which would consist of a frame-to-frame tracker and a weakly supervised tracker which would be updated under theWeakly supervised information and re-initialized by long- term detector while the trackers fail.
- Hierarchical Object Representations for Visual Recognition via Weakly Supervised Learning1
A weakly supervised approach to learn hierarchical object representations for visual recognition using bounding boxes of foreground objects as weak knowledge in the learning stage to promote those visual patterns which are more related to the target objects.
- Correlation Mining for Web News Information Retrieval1
A framework of multimodal multi-correlation news retrieval is presented, which integrates news event correlation, news entity correlation, and event-entity correlation simultaneously by exploring both text and image information.
- 1
- Active Learning for Relevance Feedback in Image Retrieval1
An SVM classifier is learned in color and texture feature subspaces, respectively and the two classifiers are used to classify the unlabeled data.
- Interactive service recommendation based on ad concept hierarchy1
This paper aims to produce interactive service recommendation based on ad concept hierarchy by linking web videos, especially ad videos, with informative product details over the commercial websites by introducing the domain based concept hierarchy.
- Sports video retargeting1
The SVR technique enables flexible and coherent aspect-ratio change of the output sports videos with a spatial-temporal 3D rectilinear grid framework, which are free from significant loss of information or distortion on salient and important regions.
- 1
- 1
- Automatic Moving Object Segmentation with Accurate Boundaries1
This paper presents a layer-model based method to segment moving objects from image sequence with accurate boundaries, using Graph Cuts technique to solve the energy minimization problem and extract motion layers.
- 1
- A framework for providing adaptive sports video to mobile devices1
A novel framework that serves adaptive sports video to mobile users that combines content-based sports highlights extraction and quality-domain video compression technologies in video server, capable of reducing wireless bandwidth consumption more effectively is presented.
- A Geometric Contour Framework with Vector Field Support1
A new geometric contour framework with support of specified vector field is proposed, and EdgeFlow, a powerful segmentation tool, is selected to generate desirable initial vector field.
- 1
- AIRE1
An ambient interactive and responsive environment (AIRE) to improve user's image browsing experiences on the small-form-factor devices by establishing an ambient interactive communication across various devices and designing a distributed interface that crosses various devices to overcome the display constraint in mobile devices.
- Face Recognition Using Overcomplete Independent Component Analysis1
The experimental results show that the overcomplete ICA can improve efficiently the recognition rate and separate out more source signals than the input data.
- Large Language Model for E-Commerce Workshop–
This workshop aims to bring together researchers and industry practitioners to explore the opportunities and limitations of LLMs in e-commerce, and foster collaboration, bridge the gap between academia and industry, and drive innovation in applying LLMs to E-Commerce.
- –
- –
- –
- –
- –
- –
- 60 Hz self-tuning background modeling–
A fast and robust approach for background modeling with self-adaptive ability that outperforms state-of-the-art approaches with a speed of 60 fps on CPU for a 640 × 480 image sequence.
- Snap & Play–
Exploring how state-of-the-art image processing techniques can assist in developing a personalized, automatic, and dynamic game shows that the proposed Personalized FiDi electronic game is satisfying in terms of player experience, seamless advertisement, and technical feasibility.
- Beyond semantic attributes: Auxiliary feature discovery for image classification–
In this paper, a Sequential augmented Attributes learning (SAL) method is proposed to implement semantic attribute augmentation, where the non-semantic mid-level features are learned one by one under a sequential error-correcting scheme so that the user can obtain more discriminating power with very compact expanded features.
- –
- Head Tracking Based on Histogram and Shape Model–
A new model-based method is presented for real-time tracking a person’s head in gray image sequences and demonstrates robustness to scale variation, arbitrary camera movement, partocclusion and so on, for several image sequences.
- –
- Social tag alignment with image regions by sparse reconstructions–
This work proposes a novel tag-to-region method with two phases of sparse reconstructions by exploring the large-scale user contributed resources to sparsely reconstruct the label vector of the given image, and uses the reconstructing weights as the semantic relevance to the image.
- –
- –
- –
- –
- Spatial pyramid based histogram representation for visual tracking with partial occlusion–
The spatial pyramid based histogram representation is introduced into the Bayesian inference framework, in which approximate global spatial information is retained and multi-scale information is explored.
- Boosted forest for human detection–
A method of boosted forest with harr-like features for human detection that associates the random decision trees as weak learners within the framework of Adaboost and dynamically combined into a strong classifier, i.e., a boosted forest.
- –
- –
- Image Segmentation Based on Supernodes and Region Size Estimation–
A kind of self-adaptive image segmentation algorithm, in which super-pixels act as the graph nodes for computational efficiency and more local features could be abstracted from the pre-segmented image.
- Region based image fusion using SVM–
This paper presents a novel fusion approach using PCA merger based on multiscale decomposition (MSD), combined with region segmentation and support vector machine (SVM), the result is a high spatial resolution multispectral image from a high resolution panchromatic (Pan) image and low resolution mult ispectral (Ms) images.
- –
- –
- –
- Representing personal web information using a topic-oriented interface–
This paper developed a novel approach to reorganize personal Web information as a topic-oriented interface to utilize anchor, title and URL information to represent content information for the browsed Web pages rather than the content body.
- Providing Adaptive Dynamic Web Content in Mobile Environment–
A novel page adaptation method called constrained page splitting is used to improve readability of dynamic pages, then split results rather than original contents are saved to cache in fragment granularity in a unified framework based on fragment-based technologies.
- ASAP: Building Multi-Client Collaborated Environments Towards Synchronized Image Sharing–
This paper proposes a novel approach called ASAP that builds multi-client collaborated environments to achieve synchronized image sharing and indicates that it is effective and useful for image sharing in mobile environment.
- Automatic Synchronized Browsing of Images Across Multiple Devices–
Experimental results show that the synchronized approach is beneficial to improve users' browsing experience, and similar images across multiple devices can be simultaneously presented for users to make comparatively viewing or searching.
- Hierarchical partition scheme in feature space for multivariate clustering–
A hierarchical partition of the feature space based on statistical information on each dimension is proposed, and then mean-shift is used to properly fuse the obtained super-cubics to reveal the genuine data structure.
- –
- Incorporating negative examples in one-class SVM for relevance feedback in image retrieval–
A new relevance feedback scheme is proposed based on the extended one-class SVM that can handle negative examples that is flexible to handle the situations with and without negative examples; and can further improve the retrieval performance when negative examples provided.
- –
- –
- –
- Mosaic Representations of Video Shots Based on Slice Image Analysis.–
An apparatus for mounting electronic components on a printed board substrate comprises a component feeder for sequentially feeding the electronic components in an aligned row, a chuck which receives and rectifies by pinching the position of the Electronic components.
- Robust Tracking of Video Objects through Topological Constraint on Homogeneous Motion–
This paper addresses the idea of area-based parametric motion estimation with spatial constraint involved, in order that the semantic segmentation and tracking of non-rigid object can be undertaken in interactive environment, which is the center demand of applications such as MPEG-4/7 or content-based video retrieval.
- A morphological method for moving object segmentation and posture recognition–
A coarse-to-fine morphological segmentation and matching method, which uses a multiscale morphological image filter and a “spore” element, which is robust enough and can be easily parallelized.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.