Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Huan WangSuggested from co-authorship
- Gang ChenSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works292 from public data
- 393
- 346
- 273
- 249
- PLBP: An effective local binary patterns texture descriptor with pyramid representation212
The conventional local binary pattern is extended to pyramid transform domain (PLBP) by cascading the LBP information of hierarchical spatial pyramids, PLBP descriptors take texture resolution variations into account and show their effectiveness for texture representation.
- 197
- Position Focused Attention Network for Image-Text Matching195
This paper proposes a novel position focused attention network (PFAN) to investigate the relation between the visual and the textual views, and integrates the object position clue to enhance the visual-text joint-embedding learning.
- Rating Prediction Based on Social Sentiment From Textual Reviews195
A sentiment-based rating prediction method (RPS) to improve prediction accuracy in recommender systems and results show the sentiment can well characterize user preferences, which helps to improve the recommendation performance.
- 134
- 131
- Learning Trajectory-Aware Transformer for Video Super-Resolution124
A novel Trajectory-aware Transformer for Video Super-Resolution (TTVSR) is proposed, which formulate video frames into several pre-aligned trajectories which consist of continuous visual tokens that enable Transformers to model long-range features.
- 103
- 103
- 100
- 98
- Preparing lessons: Improve knowledge distillation with better supervision87
Two novel approaches are introduced, Knowledge Adjustment (KA) and Dynamic Temperature Distillation (DTD), to penalize bad supervision and improve student model and when combined with other KD-based methods, the performance will be further improved.
- Long- and Short-term Preference Learning for Next POI Recommendation82
A long- and short-term preference learning model (LSPL) considering the sequential and context information of users and outperforms the state-of-art approaches for next POI recommendation.
- 81
- 81
- 79
- 78
- 78
- 76
- 74
- Text detection, localization, and tracking in compressed video73
A unified framework for text detection, localization, and tracking in compressed videos using the discrete cosines transform (DCT) coefficients is proposed and the final experimental results show the effectiveness of the proposed methods.
- 72
- 72
- 71
- Personalized location recommendation by fusing sentimental and spatial context70
A Point of Interest mining method and a personalized recommendation model by fusing sentimental spatial context and a Sentimental–Spatial POI Recommendation (SPR) model incorporating the factors of sentiment similarity and geographical distance are proposed.
- Tagging photos using users' vocabularies68
A personalized photo tagging approach by using users' own vocabularies that can recommend users preferred tags for their newly uploaded photos based on the history information in their social communities by modeling users' tagging habit is proposed.
- 67
- Recommendation via user's personality and social contextual64
Three social factors, personal interest, interpersonal interest similarity and interpersonal influence, fuse into a unified personalized recommendation model based on probabilistic matrix factorization to solve the cold start and sparsity problem of datasets.
- 4D LUT: Learnable Context-Aware 4D Lookup Table for Image Enhancement58
A novel learnable context-aware 4-dimensional lookup table (4D LUT) is proposed, which achieves content-dependent enhancement of different contents in each image via adaptively learning of photo context, and outperforms other state-of-the-art methods in widely-used benchmarks.
- 58
- 58
- 57
- Mining user-contributed photos for personalized product recommendation56
This paper proposes a novel hierarchical user interest mining (Huim) approach for personalized products recommendation, which makes full use of the visual information and UGC of its photos to mine user's interest.
- 53
- Personalized Reason Generation for Explainable Song Recommendation53
Evaluation results indicate that the generated reasons are relevant to songs and personalized to users, and they attract users to click the recommendations.
- 52
- Social media based event summarization by user–text–image co-clustering45
A new social media based event summarization framework is put forward, which comprises of three stages: a coarse-to-fine filtering model is exploited to eliminate irrelevant information, and a novel User–Text–Image Co-clustering (UTICC) is proposed to jointly discover subevents from microblogs of multiple media types—user, text, and image.
- 44
- 43
- 43
- 42
- 42
- 42
- 41
- Motion-Adaptive Separable Collaborative Filters for Blind Motion Deblurring41
A novel real-world deblurring filtering model called the Motion-adaptive Separable Collaborative (MISC) Filter, which can handle more general-ized and complex motion in a spatially differentiated man-ner and analyzes the relationships between the motion estimation network and the residual reconstruction network.
- 41
- 40
- Multisample-Based Contrastive Loss for Top-K Recommendation40
A multisample-based contrastive loss (MSCL) function that solves the two problems by balancing the importance of positive and negative samples and data augmentation by using multiple positive items simultaneously with the CL loss function.
- 40
- CaLa: Complementary Association Learning for Augmenting Comoposed Image Retrieval38
It is argued that CIR triplets encompass additional associations besides the primary query-target relation, which is overlooked in existing works and developed the CaLa, a Complementary Association Learning framework for Augmenting Composed Image Retrieval.
- 38
- 38
- Learning salient visual word for scalable mobile image retrieval38
An effective scalable mobile image retrieval approach by exploiting the advantage of mobile end that people usually take multiple photos of an object in different viewpoints and focuses, and makes full use of the multiphotos taken at mobile end to extract saliency.
- 37
- 36
- 35
- Rating prediction by exploring user’s preference and sentiment35
Three factors, including user topic, user sentiment and interpersonal influence are fused into a recommender system (RS) based on probabilistic matrix factorization, and experimental results show the proposed approach outperforms the existing approaches.
- 34
- 34
- Travel Recommendation via Author Topic Model Based Collaborative Filtering34
An Author Topic Collaborative Filtering (ATCF) method to facilitate comprehensive Points of Interest (POIs) recommendation for social media users by extracting topics about user preference from the textual description of photos by author topic model instead of from GPS.
- 33
- A review study of application of artificial intelligence in construction management and composite beams32
The performance of ANN with other soft computing techniques, such as BP, BPNN, SVR, GA, ICA, and MLFF in various fields has been compared and ANN in many cases had superiority over other models.
- Unsupervised Salient Object Detection via Inferring From Imperfect Saliency Models32
A novel salient object detection approach with the idea of inferring a superior model from a variety of previous imperfect saliency models via optimally leveraging the complementary information among them is proposed.
- HMM based soccer video event detection using enhanced mid-level semantic32
An effective hidden Markov models (HMMs) based soccer video event detection method based on a hierarchical video analysis framework is proposed and the average F-score of this approach is 82.92%, which outperforms that of DBN and CRF by 9.85% and 11.12% respectively.
- 31
- 31
- Sketch-Guided Scenery Image Outpainting31
This work proposes an encoder-decoder based network to conduct sketch-guided outpainting, where two alignment modules are adopted to impose the generated content to be realistic and consistent with the provided sketches.
- 31
- 30
- Frequency Domain-Based Diffusion Model for Unpaired Image Dehazing29
This work proposes a novel frequency domain-based diffusion model, named FrDiff, for fully exploiting the beneficial knowledge in unpaired clear data and demonstrates that it outperforms other state-of-the-art methods on both synthetic and real-world datasets.
- TTVFI: Learning Trajectory-Aware Transformer for Video Frame Interpolation29
A novel Trajectory-aware Transformer for Video Frame Interpolation (TTVFI) is proposed, which formulate the warped features with inconsistent motions as query tokens, and formulate relevant regions in a motion trajectory from two original consecutive frames into keys and values.
- 28
- 27
- Effective and efficient video text extraction using key text points27
Experimental results show that the proposed text detection and localisation algorithm is robust to the font size, style, colour and alignment of texts and decreases the influence of scene changes and motions.
- 26
- 26
- On Combining Social Media and Spatial Technology for POI Cognition and Image Localization26
A systematic review of the works that combine social media and spatial technology for POI cognition and image localization is given.
- 25
- A contrastive learning approach for ICU false arrhythmia alarm reduction25
A deep learning framework to automatically learn feature representations of physiological waveforms using convolutional neural networks (CNNs) to discriminate between true vs. false arrhythmia alarms is presented.
- Soccer Video Event Detection by Fusing Middle Level Visual Semantics of an Event Clip25
Experimental results show the effectiveness of the proposed EHMM based soccer video event detection approach, and the influences of hidden state number and overall feature types to the event detection performances are discussed.
- 24
- Kernel-attended residual network for single image super-resolution24
This work proposes a Kernel-Attended Residual Network (KARN), which possesses the optimal performance for feature fusion and feature representation and achieves a more notable performance than state-of-the-art methods by evaluating the performance of results based on benchmark datasets.
- 24
- Mining near duplicate image groups23
An automatic NDIG mining approach by utilizing adaptive global feature clustering based candidate NDIG detection and local feature refinement based NDIG verification is proposed by utilizing a hierarchical model.
- GPS Estimation from Users’ Photos23
A GPS location estimation approach is proposed for an uploaded image, its GPS information is estimated by both hierarchical global feature classification and local feature refinement to guarantee the accuracy and computational cost.
- 22
- 22
- Learning a Condensed Frame for Memory-Efficient Video Class-Incremental Learning22
This work proposes FrameMaker, a memory-efficient video class-incremental learning approach that learns to produce a condensed frame for each selected video, to reduce the memory cost by preserving only one condensed frame instead of the whole video.
- Exploring spatial and channel contribution for object based image retrieval22
This paper proposes a simple but effective method which called Strong-Response-Stack-Contribution (SRSC) to generate the global representation vector for object retrieval according to spatial and channel contribution to perform image retrieval.
- 21
- MPNET: An End-to-End Deep Neural Network for Object Detection in Surveillance Video21
This paper proposes a new framework named moving-object proposals generation and prediction framework (MPGP) to reduce the searching space and generate some accurate proposals which can reduce computational cost.
- SAR complex image data compression based on quadtree and zerotree Coding in Discrete Wavelet Transform Domain: A Comparative Study21
A novel phenomenon is observed that QC-DWT outperforms the zerotree based wavelet coding algorithms, e.g., CCSDS-IDC and Set Partitioning in Hierarchical Trees algorithm for SAR complex image data compression.
- 21
- 20
- 20
- 20
- 19
- PERD: Personalized Emoji Recommendation with Dynamic User Preference19
Experimental results validate the superiority of the proposed personalized emoji recommendation with dynamic user preference (PERD) which contains a text encoder and a personalized attention mechanism.
- Image dehazing based on a transmission fusion strategy by automatic image matting19
A transmission fusion strategy based on automatic image matting for image dehazing that significantly outperforms existing methods and can avoid the difficulty in manually labeling trimaps.
- Deep multi-task learning model for time series prediction in wireless communication19
This paper proposes a deep learning model based on LSTM for time series prediction in wireless communication, employing multi-task learning to improve prediction accuracy, and it outperforms other prediction methods.
- 19
- HWVP: hierarchical wavelet packet descriptors and their applications in scene categorization and semantic concept retrieval19
Experimental results on the widely used OT, Scene-13, Sport Event, and TRECVID 2007 datasets show the effectiveness of the proposed HWVP descriptors.
- 18
- Learning Deformable and Attentive Network for image restoration18
A Deformable and Attentive Network (DANet) to address the problems of image restoration and deformable convolution, which achieves state-of-the-art (SOTA) results for various IR tasks, including image denoising, JPEG artifacts removal, and real-world super resolution.
- 18
- 18
- Personalized Recommendation by Exploring Social Users’ Behaviors18
Two social factors: interpersonal rating behaviors similarity and interpersonal interest similarity are fused into a consolidated personalized recommendation model based on probabilistic matrix factorization and can enhance the inner link between features in the latent space.
- 17
- 17
- Detection of ships in inland river using high-resolution optical satellite imagery based on mixture of deformable part models17
A hierarchical method is proposed to detect the ships in inland river and performs particularly well in separating the ships clustered together and staying alongside the pier.
- 17
- Co-weighting semantic convolutional features for object retrieval16
An unsupervised framework that aggregates feature maps by an adaptive selection and two weighting strategies that outperforms state-of-the-art aggregation approaches by a considerable margin.
- 16
- Two-Stage Unsupervised Hyperspectral Band Selection Based on Deep Reinforcement Learning15
Experimental results demonstrated that the proposed method could effectively improve the performance of the hyperspectral image band selection and introduced the Optimal Index Factor (OIF) as the evaluation metric to improve the efficiency and quality of the selection process.
- 15
- Image classification based on low-rank matrix recovery and Naive Bayes collaborative representation15
A novel learning-free image classification algorithm under the framework of Naive-Bayes Nearest-Neighbor (NBNN) and collaborative representation, where non-negative sparse coding, low-rank matrix recovery and Collaborative representation are jointly employed to obtain more robust and discriminative representation.
- SAR image Bayesian compressive sensing exploiting the interscale and intrascale dependencies in directional lifting wavelet transform domain15
Experimental results show that the proposed DLWT Tree-Direction-Clustering Compressive Sensing (DLWT-TDC-CS) can achieve the best reconstruction performance at sampling rates from 0.5 to 0.9 compared with various state-of-the-art CS reconstruction algorithms.
- 15
- 15
- 14
- 14
- 14
- Highlight events detection in soccer video using HCRF14
An effective hidden conditional random fields (HCRFs) based soccer video event detection method is proposed and compared with the dynamic Bayesian networks (DBN) and conditionalrandom fields (CRF) based event detection approaches.
- 13
- 13
- 13
- 13
- 13
- 13
- ReGO: Reference-Guided Outpainting for Scenery Image12
ReGO (Reference-Guided Outpainting), a new method for the task of sketch-guided image outpainting that leverages neighboring reference images to synthesize texture-rich results by transferring pixels from them to facilitate texture compensating of the target image.
- 12
- Decoupling Degradations with Recurrent Network for Video Restoration in Under-Display Camera11
This paper introduces a novel video restoration network, called D2RNet, specifically designed for UDC systems, and proposes a large-scale UDC video benchmark by gathering HDR videos and generating realistically degraded videos using the point spread function measured by a commercial UDC system.
- 11
- Semi-supervised New Slot Discovery with Incremental Clustering11
This work proposes a Semi-supervised Incremental Clustering method, to discover new slots with the aid of existing linguistic annotation models and limited known slot data, and demonstrates that the proposed method significantly outperforms state-of-the-art models.
- Image super-resolution based on residually dense distilled attention network11
This paper proposes a residually dense distilled attention network (RDDAN), which is stacked in the feature extraction module of the network and achieves a more desirable performance than state-of-the-art methods.
- 11
- Improved image GPS location estimation by mining salient features11
A salient region mining and salient feature selection based image GPS estimation approach that measures the contributions of the each salient parts to GPS estimation quantitatively and fuse saliency of each BoW in determining the image taken place is proposed.
- Proportion-based robust optimization and team orienteering problem with interval data11
A proportion-based robust optimization approach is developed to deal with uncertain combinatorial optimization problems and is applied on team orienteering problem with interval data (TOPID), a variant of vehicle routing problem, which has not yet been studied before.
- 11
- 10
- 10
- Dual Relation Alignment for Composed Image Retrieval10
A new framework is proposed, termed dual relation alignment, which integrates both explicit and implicit relations to fully exploit the correlations among the triplets and is evaluated on two popular datasets, CIRR and FashionIQ.
- Enhancing performance-based generative architectural design with sketch-based image retrieval: a pilot study on designing building facade fenestrations10
The current study automates the PGAD process stepwise, making it more applicable to real design projects and tries to integrate the technology of sketch-based image retrieval (SBIR) into the selecting stage of PGAD.
- Personalized product search based on user transaction history and hypergraph learning10
A new re-ranking method for personalized product search, in which user’s transaction history is utilized to choose products which is closer to the user�'s preference into the higher positions, which is much better than the comparison methods.
- Collaboration space division in collaborative product development based on a genetic algorithm10
The genetic algorithm (GA) and minimum description length (MDL) principle may be used as tools in optimizing collaboration space, which encodes spatial information as a chromosome in binary which can be decoded into an optimal collaboration space.
- 10
- Image Retrieval by User-oriented Ranking10
A user-oriented image ranking system with the consideration of image relevance, diversity and computation complexity, aiming to automatically rank images according to their visual information, semantic information and social clues is proposed.
- 10
- Global motion estimation from randomly selected motion vector groups and GM/LM based applications10
Two applications, namely the text occluded region recovery and the error concealment, are presented using the global motion/local motion information, which shows its effectiveness and robustness against noise and motion vector loss.
- Forecasting Treatment Outcomes Over Time Using Alternating Deep Sequential Models9
Experimental results on two datasets from patients with sepsis and respiratory failure extracted from the Medical Information Mart for Intensive Care database demonstrate the effectiveness of the proposed Alternating Transformer (AL-Transformer), outperforming existing state-of-the-art methods.
- Fine-grained semantic textual similarity measurement via a feature separation network9
A feature separation network (FSN) model is proposed, which can further separate and extract attribute features and description features and then measure the semantic similarity according to the separated features.
- 9
- 9
- Compressive sensing reconstruction for compressible signal based on projection replacement9
The proposed projection replacement (PR) algorithm eliminates the hypothetic measurement error in OMP and TSW-CS reconstructed model, and it guarantees theoretically that the PR results have a smaller error.
- ISABoost: A weak classifier inner structure adjusting based AdaBoost algorithm—ISABoost based application in scene categorization9
An enhanced AdaBoost algorithm by adjusting inner structure of weak classifiers (ISABoost) is proposed and its applications in scene categorization are evaluated to show the effectiveness of the proposed algorithm.
- Video text detection and localization in intra-frames of H.264/AVC compressed video9
An overlaid text detection and localization method is proposed for H.264/AVC compressed videos by using the integer discrete cosine transform (DCT) coefficients of intra-frames to detect coarse text blocks detection using block sizes and quantization parameters adaptive thresholds.
- 8
- A Diffusion Model with Contrastive Learning for ICU False Arrhythmia Alarm Reduction8
This work proposes a deep generative approach based on the conditional denoising diffusion model to detect false arrhythmia alarms in the ICUs, and designs a network with residual links and self-attention mechanism to capture long-term dependencies in signal sequences, and leverages the contrastive learning mechanism to maximize distances between true and false alarms.
- CMID: Crossmodal Image Denoising via Pixel-Wise Deep Reinforcement Learning8
This work proposed a similarity reward to help teach an optimal action sequence to model the step-wise nature of the human processing process explicitly and designed an action set capable of handling multiple types of noise to construct the action space, thereby achieving successful crossmodal denoising.
- SAR Object Detection Encounters Deformed Complex Scenes and Aliased Scattered Power Distribution8
A high-speed and high-performance SAR image anchor-free detector that achieves the best speed-accuracy tradeoff on the SAR-ship dataset, with 96.4% average precision when the value of intersection over union is 50% ($AP_{50}$) at 64.9 frames per second.
- Rethinking Visual Relationships for High-level Image Understanding.8
A novel method to mine more valuable relationships by automatically filtering out visually-irrelevant relationships is proposed and a method to learn feature representations of instances, attributes, and visual relationships jointly from images is proposed, then applied to image captioning and visual question answering respectively.
- Deep Neural Networks for Free-Hand Sketch Recognition8
This paper proposes a CNN training on contours that performs well on sketch recognition over different databases of the sketch images and makes some adjustments to the contours for training to reach higher recognition accuracy.
- LCMKL8
This paper proposes a novel approach called latent-community and multi-kernel learning (LCMKL), which outperforms other state-of-the-art approaches on NUS-WIDE database.
- Scale and rotation invariant Gabor texture descriptor for texture classification8
This paper presents a scale and rotation invariant Gabor Texture Descriptor (GTD) for texture classification that combines Gabor filters with different orientations and scales and a new method using Discrete Fourier Transform on the scale invariant GTD.
- An SVM-Based Soccer Video Shot Classification Scheme Using Projection Histograms8
An SVM based shot classification scheme, which categorizes shots of soccer videos into 4 kinds, namely global shots, medium shots, close-up shots and audience shots, is proposed, which performs competitively and a novel feature, the projection histogram, is effective.
- 8
- 8
- 7
- Learning Data-Driven Vector-Quantized Degradation Model for Animation Video Super-Resolution7
A multi-scale Vector-Quantized Degradation model for animation video Super-Resolution (VQD-SR) is proposed to decompose the local details from global structures and transfer the degradation priors in real-world animation videos to a learned vector-quantized codebook for degradation modeling.
- Towards Handling Sudden Changes in Feature Maps During Depth Estimation7
A Compressive Sensing based Depth Representation (CSDR) scheme is proposed, which formulates the problem of depth estimation in pixel space into the task of fixed-length vector regression in representation space, and claims that regression in CSDR space performs better than traditionally direct depth map estimation in pixels space.
- Dialogue State Tracking Based on Hierarchical Slot Attention and Contrastive Learning7
To address the problem that the dialogue utterance is semantically distant from the corresponding slot value, the contrastive learning is introduced to make the utterance embedding mapped under each slot name more suitable with the ground truth value and away from other slot values, which improves the accuracy of mapping between feature space and semantic space.
- 7
- 7
- Recommend Social Network Users Favorite Brands7
A novel brands recommendation approach for social network users based on their browsing images and labeled texts is proposed, which enrich the uploaded image's texts by image annotation approach, and builds brand tree from the collected datasets.
- 7
- 7
- 6
- AMP-BCS: AMP-based image block compressed sensing with permutation of sparsified DCT coefficients6
An approximate message passing (AMP)-based BCS (AMP-BCS) method that significantly outperforms current state-of-the-art BCS algorithms in both visual perception and objective metrics is proposed.
- 6
- 6
- 6
- Adaptive Co-Weighting Deep Convolutional Features for Object Retrieval6
An efficient unsupervised aggregation method that uses an adaptive Gaussian filter and an element-value sensitive vector to co-weight deep features to improve the discrimination power of image vectors is proposed.
- 6
- Hyperspectral image compression based on DLWT and PCA6
This paper first uses principal component analysis (PCA), then a directional lifting wavelet transform (DLWT) is used to remove the spectral correlation, and the proposed image compression scheme achieves higher performances when compared with DWT based Consultative Committee for Space Data Systems (CCSDS).
- 6
- 6
- Hidden conditional random field-based soccer video events detection6
HCRF is utilised to model the observations of mid-level semantics of an event clip for event detection and Comparisons are made with the dynamic Bayesian networks, hidden Markov model (HMM), enhanced HMM and conditional random field-based event detection approaches.
- Boosted Scene Categorization Approach by Adjusting Inner Structures and Outer Weights of Weak Classifiers6
The inner structures of weak classifiers are adjusted before their outer weights determination and Experimental results on three AdaBoost algorithms show the effectiveness of the proposed approach.
- Graph-cut based tag enrichment6
Experiments on Flickr dataset demonstrate the effectiveness of the proposedGraph cut based tag enrichment approach, which builds a graph for each image with its initial tags.
- M3Net: Efficient Time-Frequency Integration Network with Mirror Attention for Audio Classification on Edge5
This work introduces a novel multi-view method for the separated time-frequency features extraction and utilization, which exists within the proposed Mini Mirror Multi-View Network (M3Net) in the form of the Mirror Attention mechanism, which enables reversible spatial transformation of spectral features.
- QueryCDR: Query-Based Controllable Distortion Rectification Network for Fisheye Images5
A novel Query-Based Controllable Distortion Rectification network for fisheye images (QueryCDR), which defines the latent spatial relationships for different distortion degrees as a series of learnable queries and proposes two kinds of controllable modulating blocks to enable the control conditions to guide the modulation of the distortion features better.
- 5
- Social image retrieval based on topic diversity5
A novel method to diversify the retrieval results by latent topic analysis by employing NMF (Non-negative Matrix Factorization) and an adaptive multi-feature fusion model to learn the final relevance score is proposed.
- 5
- Spatial Verification for Scalable Mobile Image Retrieval5
The proposed mobile image retrieval algorithm first determines the relevant photos according to visual similarity in mobile end, then mines salient visual words by exploring saliency from multiple relevant images, and finally it determines the contribution order of salient visual Words for scalable retrieval.
- 5
- An effective GM/LM-based video error concealment5
Experimental results show the effectiveness of the proposed GM/LM-based error concealment method, which separates correct-macro-blocks into global motion compensated MBs and local motion compensatedMBs adaptively.
- Learning Deblurring Texture Prior From Unpaired Data with Diffusion Model4
TP-Diff provides a promising unsupervised deblurring solution and outperforms SOTA methods in widely-used benchmarks and implements a waveletbased adversarial loss to preserve high-frequency texture details.
- Ear with Eye: Lightweight Multimodal Audio-Visual Network Inspired by Bionic Structures4
Experimental findings show that BIMNet achieves superior performance and efficiency in diverse image datasets, audio datasets (encompassing various types of human and environmental sound), and audio-visual joint event detection tasks.
- 4
- 4
- Pseudo-Label Enhanced Prototypical Contrastive Learning for Uniformed Intent Discovery4
A Pseudo-Label enhanced Prototypical Contrastive Learning (PLPCL) model for uniformed intent discovery is proposed and a prototype learning method integrating the supervised and pseudo signals from IND and OOD samples is designed.
- 4
- Image Taken Place Estimation via Geometric Constrained Spatial Layer Matching4
A global feature clustering and local feature refinement based image location estimation approach that builds “spatial layers of visual word” (SLW) is built as an extension of the unorganized bag-of-words image representation.
- Global Motion Estimation and Its Applications4
Four global motion based applications in video compression, sport video shot classification, video error concealment, and video text occluded region recovery are given.
- 4
- Sequence-augmented Conversational Recommendation System Based on Diffusion Models for Personalized Cultural Exploration3
SDCRS is proposed, a sequence-augmented CRS based on diffusion models, which integrates sentence-level and entity-level sequential modeling to enhance the response generation and recommendation modules and improves the diversity of generated responses but also enhances user preference modeling.
- 3
- Discard Significant Bits of Compressed Sensing: A Robust Image Coding for Resource-Limited Contexts3
This research indicates that leveraging the image prior enables the estimation of most significant bits from least significant bits, which provides a quantization strategy to improve R-D performance without increasing coding complexity.
- 3
- 3
- Dual Path Network for Cross-view Geo-Localization3
This work reevaluate the square-ring partition strategy proposed in LPN and identifies a limitation in its fixed splitting strategy of features, and proposes a multi-scale ring partition strategy that effectively partitions the features at different scales to accommodate the change in the target scale.
- Directional lifting wavelet transform domain image steganography with deep-based compressive sensing3
This work proposes a novelty image steganography via deep-based compressive sensing for the reconstructed image and directional lifting wavelet transform (DLWT) for the stego image that can achieve visual quality, robustness, and security while maintaining the security of the system.
- A Non-Local Enhanced Network for Image Restoration3
This paper proposes a batch-wise non-local module to explore richer non- local dependencies within images and builds the Non-local Enhanced Network (NLENet), which achieves state-of-the-art performance on several IR benchmark datasets.
- Optimization algorithms for composite beam as smart active control of structures using genetic algorithms3
It has been found that an effective model of smart composite plates can predict the dynamic characteristics and GA could show better performance for managing linear feedback laws under given assumptions.
- Multi-source fusion based geo-tagging for web images3
An effective image GPS (geo-coordinates or geo-tags) estimation approach is proposed by fusing the multi-source such as textual, temporal and visual features of web images to inference the GPS of web image.
- 3
- Product Image Search with Deep Attribute Mining and Re-ranking3
This paper explores deep learning with convolutional neural networks to resolve query’s classification, and proposes an efficient approach for product image search that significantly outperforms state of art in product imageSearch.
- Spatial Constraint for Image Location Estimation3
A global feature clustering and local feature refinement based image location estimation approach that exploits spatial information by processing useful visual words and generates a position descriptor for each visual word.
- Video content categorization using the double decomposition3
A new framework for video content modeling is proposed, under which video contents are decomposed into multiple interacting processes by double decomposition that aims at each type of combination of structures.
- Scene Categorization Using Boosted Back-Propagation Neural Networks3
Experimental results on the widely used Scene-13 and Sport Event datasets and Comparisons with some existing scene categorization methods using the authors' datasets further show effectiveness of the proposed boosted BPN based approach.
- 2
- CanonFedRec: A Canonical Geometric Framework for Personalized Federated Recommendation2
CanonFedRec is proposed, a novel framework that achieves superior performance while reducing memory costs by up to 40× compared to the best FedRec approach and treats item-wise variance as a proxy for client-side cognitive divergence, and dynamically adapting optimization objectives via elastic decision boundaries.
- 2
- 2
- 2
- 2
- 2
- 2
- Semantic Gated Network for Efficient News Representation2
This paper develops a novel framework, namely Semantic Gated Network (SGN), to integrate the news title, tags and visual illustrations to obtain an efficient joint textual-visual feature for the news, by which one can directly measure the relevance between two pieces of news.
- Social Image Aesthetic Measurement based on 3D Reconstruction2
A model that employs some heuristic principles for photographing in three-dimensional space and based on a set of composition guidelines, including Dynamic Balance, Rule of Right Thirds and Diagonal Dominance, the composition aesthetics of each photo is estimated.
- An Approach to the Compact and Efficient Visual Codebook Based on SIFT Descriptor2
Experiments prove that application of the proposed compact codebook not only reduces computational complexity, but also improves performance of object classification.
- 1
- 1
- 1
- 1
- A Plug-and-Play Model-Agnostic Embedding Enhancement Approach for Explainable Recommendation1
RVRec is introduced, a plug-and-play model-agnostic embedding enhancement approach that can improve both personality and explainability of existing systems and introduces a probability-based embedding optimization method that uses a contrastive loss based on negative 2-Wasserstein distance to enhance the representativeness of the embeddings.
- Knowledge Transfer-Driven Few-Shot Class-Incremental Learning1
Extensive quantitative and qualitative experiments on three popular FSCIL benchmark datasets demonstrate that the proposed method, named Knowledge Transfer-driven Relation Complementation Network (KT-RCNet), outperforms almost all prior methods.
- Structural Design and Dynamics Simulation of Lower Limb Fitness Exoskeleton1
The fitness exoskeleton structure of lower limbs was designed for exercising people's lower limb muscles and cardiorespiratory function by combining the functions, operating principle, and structural composition implemented by the exoskeleton.
- Logic Diffusion for Knowledge Graph Reasoning1
This work proposes a plug-in module called Logic Diffusion (LoD) to discover unseen queries from surroundings and achieves dynamical equilibrium between different kinds of patterns and proves the general effectiveness of LoD on the noise-rich knowledge graph.
- Scaled gated networks1
A novel scaled gated convolution is proposed that enables attention-enhanced CNNs to overcome the paradox between performance and redundancy and can be easily applied to modern CNNs in a plug-and-play manner.
- 1
- 1
- Scale and Topology Preserving SIFT Feature Hashing1
This work proposes a hashing scheme called Geometry and Topology Preserving Hashing for content based image retrieval that gives better precision in experiment on the Oxford Building dataset.
- Semantic Based Sport Video Browsing1
In this chapter, this chapter focuses its attention on semantic based sport video highlights detection and semanticbased sport video browsing.
- A Video Text Detection Method Based on Key Text Points1
Experimental results show that the proposed text detection method is robust to the font size, style, color and alignment of texts, and greatly speeds up the text detection.
- Finite Element Analysis of the Spindle of CA6140 Lathe1
The spindle model of CA6140 is set up and by analyzing the model with Finite Element method, the paper put forward several useful ways to reduce the machining errors.
- 1
- –
- –
- –
- –
- –
- JSRC-diff: a diffusion-based joint framework for hyperspectral image super-resolution and classification–
The Spatial-Spectral Decoupled U-Net is constructed as the backbone for diffusion denoising, reconstructing high-resolution HSI during diffusion iterations while simultaneously outputting multi-scale features and introducing a Multi-Temporal Multi-Layer Feature Fusion (MTMLF) classification module during diffusion iterations.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Exploring Spectral Characteristics for Single Image Reflection Removal–
This paper proposes the Spectral Codebook to reconstruct the optical spectrum of the reflection image, and designs two spectral prior refinement modules to re-distribute pixels in the spatial dimension and adaptively enhance the spectral differences along the wavelength dimension.
- –
- –
- –
- Federated Consistency- and Complementarity-aware Consensus-enhanced Recommendation–
The proposed Fed3CR is a plug-and-play method, which can be integrated with other FedRec methods to improve its performance, and enables the client to adaptively enhance specific information in the consensus, transforming it into a form that best suits itself.
- –
- –
- –
- Motion and Trajectory Planning and Simulation of a High Frequency Transformer Winding Robot Pose–
An optimization trajectory optimization algorithm based on the combination of accurate unit scanning and Cartesian spatial trajectory planning for smooth processing of high-frequency transformers is proposed.
- –
- –
- –
- –
- –
- Evaluating the performance of free-formed surface parts using an analytic network process–
A Hybrid ANP (HANP) structure for evaluating the part's curved surface is constructed and a HANP evaluation of an impeller is presented to illustrate the application of the proposed method.
- Compressive imaging based on coefficients cutting in DLWT domain and approximate message passing–
A compressive imaging scheme based on coefficients cutting in directional lifting wavelet transform (DLWT) domain and low-complexity iterative Bayesian algorithm is proposed, which offers better recovery performance with much lower complexity than the previous work.
- A fine-grain nonlocal weighted average method for image CS reconstruction–
This paper proposes a novel fine-grain nonlocal weighted average method for nature image CS reconstruction, and takes full use of the nonlocal repetitive structures to recover image from the observations to outperform the existing classical CS reconstruction algorithms in Peak-Signal-Noise-Ratio (PSNR) and subjective evaluation.
- A Face Retrieval Method Based on Resampling of Interested Region–
The approach realizes the weighting of important region in face retrieval, and its performance is dramatically enhanced under the circumstance of little increase of computational complexity, compared with the existing uniformed blocking method.
- An Efficient Global Motion Estimation Algorithm on H.264/AVC Compression Domain–
The noise blocks were eliminated efficaciously by using information of MVD and reference block for MV prediction in H.264 bits stream and the proposed method is efficient and robust in terms of both computational complexity and accuracy.
- An improved fusion method based on adaboost algorithm for semantic concept extraction–
Based on probability distribution of weak classifier output, an improved Adaboost-based multi-classifiers fusion algorithm is proposed for semantic concept extraction by presenting a novel method to compute the error rate and the weight of each classifier.
- –
- A Triangular-Mesh Based Methodology of Model Description in Collaborative Design and Prototype Manufacturing of Surface–
The infrastructure uses a method based on triangular mesh model to descript the surface and meets the requirements well in collaborative design and rapid prototype manufacturing of surface.
- –
- –
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.