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
- Zhaoxuan YangSuggested from co-authorship
- Jintao LiSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works364 from public data
- 390
- 369
- Focus Your Attention213
A novel Bidirectional Focal Attention Network (BFAN) is presented, which not only allows to attend to relevant fragments but also diverts all the attention into these relevant fragments to concentrate on them, which enables to avoid preference to long text or complex image.
- 133
- 127
- 124
- 119
- 116
- 112
- 112
- Disentangled Multimodal Representation Learning for Recommendation108
This paper proposes a novel Disentangled Multimodal Representation Learning (DMRL) recommendation model, which can capture users' attention to different modalities on each factor in user preference modeling, and employs a disentangled representation technique to ensure the features of different factors in each modality are independent of each other.
- 97
- Scene graph captioner: Image captioning based on structural visual representation97
This work proposes a novel framework to embed a scene graph into the structural representation, which captures the semantic concepts and the graph topology and develops the scene-graph-driven method to generate the attention graph.
- Behavior pattern clustering in blockchain networks96
This paper introduces the problem of behavior pattern clustering in blockchain networks and proposes a novel algorithm termed BPC for this problem, evaluating a long list of potential sequence similarity measures, and selecting a distance that is suitable for the behaviorpattern clustering problem.
- 95
- 93
- 88
- Coupled hidden conditional random fields for RGB-D human action recognition88
The extensive experiments show that the proposed model can boost the performance of human action recognition by taking advance of complementary characteristics from both RGB and depth modalities.
- 87
- CAT-DM: Controllable Accelerated Virtual Try-On with Diffusion Model82
A basic diffusion-based virtual try-on network is designed, which utilizes ControlNet to introduce additional control conditions and improves the feature extraction of garment images and demonstrates the superiority of CAT-DM against both GAN-based and diffusion-based methods in producing more real-istic images and accurately reproduced garment patterns.
- 78
- 78
- 77
- 72
- Graph-based characteristic view set extraction and matching for 3D model retrieval70
A novel method for characteristic view extraction and similarity measurement is proposed using the graph clustering method for view grouping and the random-walk algorithm is applied to adaptively update the weight of each view.
- 67
- 67
- 65
- 60
- 60
- 59
- 58
- Single/multi-view human action recognition via regularized multi-task learning58
The pyramid partwise bag of words (PPBoW) representation is proposed which implicitly encodes both local visual characteristics and human body structure and can achieve competing performance against the state-of-the-art methods for human action recognition on KTH and MV-TJU.
- Phase contrast time-lapse microscopy datasets with automated and manual cell tracking annotations57
This comprehensive, validated dataset will be useful in advancing the development of computer-aided cell tracking algorithms and function as a benchmark, providing an invaluable opportunity to deepen the understanding of individual and population-based cell dynamics for biomedical research.
- 56
- Hierarchical multi-view context modelling for 3D object classification and retrieval56
A hierarchical multi-view context modelling method for 3D object classification and retrieval that can imitate the human need to look back and forth to identify and compare the discriminative parts of individual 3D objects based on a joint convolutional neural network and bidirectional long short-term memory network.
- Spatiotemporal Joint Mitosis Detection Using CNN-LSTM Network in Time-Lapse Phase Contrast Microscopy Images52
A framework to aid humans in annotating mitosis with high efficiency and accuracy in raw phase contrast microscopy images based on the joint detection results using the proposed method is developed.
- 49
- 48
- 48
- 46
- Dual-level Embedding Alignment Network for 2D Image-Based 3D Object Retrieval46
This is the first unsupervised work to jointly realize cross-domain feature learning and distribution alignment in an end-to-end manner for this new task, 2D image-based 3D object retrieval.
- BooW-VTON: Boosting In-the-Wild Virtual Try-On via Mask-Free Pseudo Data Training45
BooW-VTON is introduced, the mask-free virtual try-on diffusion model, which delivers SOTA try-on quality without parsing cost and ensures the model’s mask-free try-on capability by creating high-quality pseudo-data and enhances its handling of complex spatial information through effective in-the-wild data augmentation.
- Image-Based Virtual Try-On: A Survey45
This survey provides a comprehensive analysis of the state-of-the-art techniques and methodologies in aspects of pipeline architecture, person representation and key modules such as try-on indication, clothing warping and try-on stage in image-based virtual try-on.
- 45
- 44
- 43
- 42
- 40
- Surface Labeling of Enveloped Viruses Assisted by Host Cells40
Since the viral natural assembly process was employed for the labeling, potential threats of genetic engineering and difficulties in keeping viral natural bioactivity were avoided and the technical complexity greatly reduces.
- 38
- 38
- LSTM-based multi-label video event detection38
A novel architecture for jointly recognizing multiple events in a given surveillance video, motivated by the sequence to sequence network is proposed, which can predict what happens in a video directly without the preprocessing of object detection and tracking.
- Multi-modal Correlated Network for emotion recognition in speech37
A novel Multi-modal Correlated Network for emotion recognition aiming at exploiting the information from both audio and visual channels to achieve more robust and accurate detection and demonstrates the correlation information of multiple modals is crucial for automatic emotion recognition.
- 37
- Single/cross-camera multiple-person tracking by graph matching37
This work is the first to unify two kinds of tracking problems with the same framework by graph matching, and is mathematically formulated into the Rayleigh Quotients Maximization.
- Hierarchical & multimodal video captioning: Discovering and transferring multimodal knowledge for vision to language36
A Hierarchical & Multimodal Video Caption (HMVC) model is proposed to jointly learn the dynamics within both visual and textual modalities for video caption task, which infers an arbitrary length sentence according to the input video with arbitrary number of frames.
- 36
- 35
- Label Semantic Knowledge Distillation for Unbiased Scene Graph Generation34
A novel model-agnostic Label Semantic Knowledge Distillation (LS-KD) is proposed for unbiased SGG, which dynamically generates a “soft” label for each subject-object instance by fusing a predicted Label Semantic Distribution (LSD) with its original one-hot target label.
- Median filtering forensics in digital images based on frequency-domain features34
Experimental results obtained on widely used databases show that the proposed method achieves outstanding performance in distinguishing median-filtered images from original images or images that have undergone other types of manipulations, demonstrating its great advantage to be applied in real-time processing of big multimedia data.
- Human Action Recognition Via Multi-modality Information34
The proposed pyramid appearance and global structure action descriptors on both RGB and depth motion history images and a model-free method for human action recognition outperforms the state-of-the-art methods on both MSR Action3D and DHA datasets.
- 33
- SP-VITON: shape-preserving image-based virtual try-on network33
This work proposes an SP-VITON (Shape-Preserving VIrtual Try-On Network) to keep the user’s original body shape while getting rid of the original clothes.
- Multiview and Multimodal Pervasive Indoor Localization33
This paper addresses the deficiency of using a single measurement by proposing the original idea of localization based on multi-view and multi-modal measurements, and constructs a benchmark dataset for the PIL task, which to this knowledge is the first public dataset that is tailored for multi-View multi- modal indoor localization and contains both magnetic and visual signals.
- 32
- Photocatalytic Dehydrogenation of Primary Alcohols: Selectivity Goes against Adsorptivity31
It is reported that the selectivity of photocatalytic dehydrogenative oxidations of aliphatic primary alcohols in acetonitrile solution into corresponding aldehydes exhibits an anomalous relationship with adsorption behavior of the alcohols.
- 30
- Smooth filtering identification based on convolutional neural networks30
A new approach based on convolutional neural networks (CNNs) is proposed, where distinguishable frequency-domain features are put into a conventional CNN model, to identify the template parameters of various types of spatial smooth filtering operations, such as average, Gaussian and median filtering.
- 3D object retrieval based on sparse coding in weak supervision30
This paper applies the sparse coding method in a weakly supervision manner to address 3D model retrieval and uses FDDL to learn dictionary to reconstruct query sample for retrieval.
- 29
- 29
- MMJN: Multi-Modal Joint Networks for 3D Shape Recognition29
A novel multi-modal joint network (MMJN) for 3D shape recognition is proposed, which can consider the correlation between two different modalities to extract the robust feature vector and define the weight of differentmodalities features based on the statistic method and utilize the advantages of different modality to generate more robust feature.
- Multi-Level Policy and Reward Reinforcement Learning for Image Captioning29
This work proposes a novel multi-level policy and reward RL framework for image captioning that contains a Multi-Level Policy Network that can adaptively fuse the word- level policy and the sentence-levelpolicy for the word generation and a guidance term to bridge thepolicy and the reward for RL optimization.
- 29
- Multi-modal & Multi-view & Interactive Benchmark Dataset for Human Action Recognition29
This paper introduces a Multi-modal & Multi-view & Interactive (M2I) dataset, which is designed for the evaluation of the performances of human action recognition under multi-view scenario, and evaluates three representative methods for the single-view, cross view, and cross domain humanaction recognition on this dataset.
- 28
- 28
- 27
- Self-Attention Graph Residual Convolutional Networks for Event Detection with dependency relations27
A novel model Self-Attention Graph Residual Convolution Networks (SAGRCN) is proposed to mine node-to-node latent dependency relations via self-attention mechanism and introduce GraphResidual Network (GResNet) to solve graph information vanishing problem.
- Semantic Consistency Guided Instance Feature Alignment for 2D Image-Based 3D Shape Retrieval27
A novel semantic consistency guided instance feature alignment network (SC-IFA), which designs two feature translators to transform the feature from one domain to another domain, and imposes the translation loss and correlation loss on the transformed features to preserve the semantic consistency.
- HGAN27
A novel Holistic Generative Adversarial Network (HGAN) is proposed to solve the cross-domain feature representation problem and make the feature space of virtual characteristic view more inclined to thefeature space of the real picture.
- 26
- Convolutional deep learning for 3D object retrieval26
A feature extraction model based on convolutional neural networks (CNN) that extracts a set of 2D images from 3D model to represent each 3D object and nearest neighbor is used to compute the similarity between different 3D models in order to handle the retrieval problem.
- 25
- Multi-Year ENSO Forecasts Using Parallel Convolutional Neural Networks With Heterogeneous Architecture25
Experimental results demonstrate that the proposed novel parallel deep convolutional neural network (CNN) with a heterogeneous architecture outperforms the other well-known methods, especially for long-term predictions.
- 25
- Hierarchical Deep Neural Network for Image Captioning25
A hierarchical deep neural network is proposed, which consists of the bottom layer and the top layer, which extracts the visual and high-level semantic information from image and detected regions, respectively, while the latter integrates both of them with adaptive attention mechanism for the caption generation.
- MV-CLIP: Multi-View CLIP for Zero-Shot 3D Shape Recognition24
This paper introduces view selection in the vision side that minimizes entropy to identify the most informative views for 3D shape and proposes hierarchical prompts combined of hand-crafted and GPT-generated prompts to refine predictions.
- 24
- Hierarchical Instance Feature Alignment for 2D Image-Based 3D Shape Retrieval24
HIFA consists of two modules, cross-modal instance feature learning and hierarchical instance feature alignment, and extensive experiments validate the superiority of HIFA for 2D image-based 3D shape retrieval task.
- A Fine-Grained Spatial-Temporal Attention Model for Video Captioning24
The proposed FSTA model achieves the spatial-hard attention at a fine-grained region level of objects through the mask pooling module and compute the temporal soft attention by using a two-layer LSTM network with attention mechanism to generate sentences.
- Cross-Domain 3D Model Retrieval via Visual Domain Adaption24
The proposed method can significantly outperform the most recent method for cross-domain 3D model retrieval and the champion of Shrec’16 Large-Scale 3D Shape Retrieval from ShapeNet Core55.
- 23
- 22
- 22
- Improved Semantic Representation Learning by Multiple Clustering for Image-Based 3D Model Retrieval22
This paper first extracts features for 2D images and 3D models, then clusters combining the 3D features with the semantic information from multiple clustering on 3D model features to obtain more reliable target pseudo label.
- Cross-view action recognition by cross-domain learning22
A novel cross-domain learning method to handle action recognition by discovering and sharing common knowledge among different video sets captured in multiple viewpoints and applying the block-wise weighted kernel function to leverage cross-view information.
- Evaluation of local spatial–temporal features for cross-view action recognition22
This paper evaluates four popular spatial-temporal features (STIP, Cuboids, MoSIFT, HoG3D) with the framework of transferable dictionary pair learning and demonstrates the superiority of this framework on cross-view human action recognition.
- MCDAN: A Multi-Scale Context-Enhanced Dynamic Attention Network for Diffusion Prediction21
It is argued that user intent understanding is also a key part of information diffusion prediction, and a novel Multi-scale Context-enhanced Dynamic Attention Network (MCDAN) is proposed to predict which user will most likely join the observed current cascades.
- Multi-Modal Validation and Domain Interaction Learning for Knowledge-Based Visual Question Answering21
- Towards Deconfounded Image-Text Matching with Causal Inference21
The Deconfounded Causal Inference Network (DCIN) is proposed, which decomposes the intra- and inter-modal confounders and incorporates them into the encoding stage of visual and textual features, effectively eliminating the spurious correlations during image-text matching, and uses causal inference to mitigate biases of external knowledge.
- Dilated Convolution-based Feature Refinement Network for Crowd Localization21
This work proposes a novel Dilated Convolution-based Feature Refinement Network (DFRNet) to enhance the representation learning capability and introduces a Feature Perception Module to model long-range contextual information at different scales by adopting multiple dilated convolutions.
- 21
- Dynamic Causal Disentanglement Model for Dialogue Emotion Detection20
A Dynamic Causal Disentanglement Model is proposed, which effectively decomposes the content of dialogues and investigates the temporal accumulation of emotions, thereby enabling more precise emotion recognition.
- 20
- 20
- 20
- Nonnegative Mixed-Norm Convex Optimization for Mitotic Cell Detection in Phase Contrast Microscopy20
This method can overcome the difficulty in feature formulation for deformable objects and is independent of tracking or temporal inference model and can produce competing results with the state-of-the-art methods.
- 19
- 19
- 19
- 18
- 18
- 18
- 18
- 18
- Multi-View Hierarchical Fusion Network for 3D Object Retrieval and Classification18
Experimental results show that the proposed multi-view hierarchical fusion network (MVHFN) outperforms the current state-of-the-art methods in both the 3D object retrieval and classification tasks.
- SHREC'18: RGB-D object-to-CAD retrieval18
Evaluation results show that while the RGB-D to CAD retrieval problem is indeed challenging due to incomplete RGB-D reconstructions, it can be addressed to a certain extent using deep learning techniques trained on multi-view 2D images or 3D point clouds.
- RGB-D Object-to-CAD Retrieval18
Evaluation results show that while the RGB-D to CAD retrieval problem is indeed challenging due to incomplete RGB-D reconstructions, it can be addressed to a certain extent using deep learning techniques trained on multi-view 2D images or 3D point clouds.
- 18
- 18
- TRCE: Towards Reliable Malicious Concept Erasure in Text-to-Image Diffusion Models17
This study proposes TRCE, using a two-stage concept erasure strategy to achieve an effective trade-off between reliable erasure and knowledge preservation, and demonstrates its effectiveness in erasing malicious concepts while better preserving the model's original generation ability.
- Multiscale Spatio-Temporal Attention Network for Sea Surface Temperature Prediction17
A novel multi-scale spatio-temporal attention network, named MUSTAN, tailored for the SST prediction problem, which achieves multi-scale fusion through a progressive scale expansion paradigm, enabling the propagation of fine-scale SST changes across broader scales.
- 17
- 17
- 17
- HEp-2 cells Classification via clustered multi-task learning17
The extensive experiments on the Hep 2, cell dataset released by the HEp-2 Cells Classification contest show that the proposed method can discover and share the latent relatedness among multiple tasks and consequently augment the performance.
- 16
- 16
- 16
- Part-Aware Interactive Learning for Scene Graph Generation16
A part-aware interactive learning method that can highlight the subject-related object parts and the object-related subject parts during inter-object interactive learning and demonstrate the superiority of the method against the state of the arts on Visual Genome is proposed.
- 16
- 16
- 16
- Structure Causal Models and LLMs Integration in Medical Visual Question Answering15
This work proposes a causal inference framework for the MedVQA task, which effectively eliminates the relative confounding effect between the image and the question to ensure the precision of the question-answering (QA) session, and introduces a novel causal graph structure that represents the interaction between visual and textual elements.
- 15
- 15
- Generic Image Manipulation Localization through the Lens of Multi-scale Spatial Inconsistence15
The proposed end-to-end method to identify the multi-scale spatial inconsistency for image manipulation localization can outperform state-of-the-art methods by a significant margin in terms of average F1 score while maintaining robustness with respect to various attacks.
- Semantic Completion and Filtration for Image–Text Retrieval15
A novel semantic completion and filtration (SCAF) method to generate a complete semantic description of an image using multi-view text descriptions, guiding the model to explore the correlations between relevant region–word pairs fully.
- Mask and Predict15
A novel multi-step reasoning manner for SGG is proposed, which breaks SGG into two explicit learning stages, including intuitive training stage (ITS) and rationalTraining stage (RTS), yielding an intuitive scene graph.
- Triangle-Reward Reinforcement Learning15
This work argues that the XE objective is not sensitive to visual-linguistic alignment, which cannot discriminately penalize the semantic inconsistency and shrink the context gap, and proposes the Triangle-Reward Reinforcement Learning (TRRL) method, which demonstrates the superiority of TRRL on the MSCOCO dataset.
- Comprehensive image quality assessment via predicting the distribution of opinion score15
This work focuses on the distribution of opinion scores (DOS) and aims to comprehensively evaluate image quality via automatically predicting DOS and is the first to introduce label distribution learning approach for image quality assessment.
- Cross-domain semantic transfer from large-scale social media15
A cross-domain semantic modeling method for automatic image annotation for visual information from social network platforms, which leverages cross- domain datasets to discover the common knowledge of each semantic concept from UGCs and boost the performance of semantic annotation by semantic transfer.
- Spatiotemporal Sparsity Induced Similarity Measure for Human Action Recognition15
A model-free method for human action recognition via sparse spatiotemporal representation that can handle the action recognition under difference scenarios by multiple persons, and which outperforms most of the state-of-art methods for human behavior recognition.
- Personalized movie recommendation15
The promising results of subjective evaluation indicate that the proposed framework can discover the latent relationship between movies as well as movie highlights and therefore provide personalized movie recommendation to effectively lead movie access in an individualized manner.
- Better Fit: Accommodate Variations in Clothing Types for Virtual Try-On14
An adaptive mask training paradigm that dynamically adjusts training masks is proposed that improves the alignment and fit of clothing but also significantly enhances the fidelity of virtual try-on experience and two metrics for unpaired try-on evaluation, the Semantic-Densepose-Ratio (SDR) and Skeleton-LPIPS (S-LPIPS) are proposed.
- 14
- 14
- 14
- PAGN: perturbation adaption generation network for point cloud adversarial defense14
A novel perturbation adaption generation network for point cloud adversarial defense, which can adaptively generate adversarial point cloud samples according to the point cloud to improve the robustness of the deep learning model for the 3D model classification.
- Joint deep feature learning and unsupervised visual domain adaptation for cross-domain 3D object retrieval14
This paper proposes a joint deep feature learning and visual domain adaptation method (Deep-VDA) to solve the cross-domain 3D object retrieval problem by the end-to-end learning, benefiting from the advantages of deep learning networks.
- LCC: Learning to Customize and Combine Neural Networks for Few-Shot Learning14
This work aims to meta-learn how to effectively combine several base-learners and proposes to learn not only a single base-learner but an ensemble of several base-learners to obtain more robust results.
- 2D Scene Sketch-Based 3D Scene Retrieval14
The objective of this track is to evaluate the performance of different 2D scene sketch-based 3D scene retrieval algorithms using a 2D sketch query dataset and a 3D Warehouse model dataset and suggest several future research directions to tackle this research problem.
- Subgraph learning for graph matching14
A novel approach is proposed to handle graph matching problem based on Markov Chain Monte Carlo framework that can effectively save matching pairwise under one-to-one matching constraints and also avoid the effect of outliers and deformation.
- 14
- Jointly Learning Multiple Sequential Dynamics for Human Action Recognition14
This work theoretically proposes the multi-task conditional random fields model and explores its application on human action recognition, and proposes the part-induced spatiotemporal action unit sequence to represent each action sample with multiple partwise sequential feature subspaces.
- 14
- MarkPlugger: Generalizable Watermark Framework for Latent Diffusion Models Without Retraining13
This work proposes MarkPlugger, a generalizable plug-and-play watermark framework without LDM retraining that effectively harmonizes image quality and watermark recovery rate and is generalized to multiple official versions and modified variants of LDMs, even without retraining the watermark model.
- 13
- Revealing Vulnerabilities in Stable Diffusion via Targeted Attacks13
This study designs a gradient-based embedding optimization method to craft reliable adversarial prompts that guide stable diffusion to generate specific images and reveals the mechanisms that cause the vulnerability of the model.
- 13
- My Brother Helps Me: Node Injection Based Adversarial Attack on Social Bot Detection13
This study proposes a node injection-based adversarial attack method designed to deceive bot detection models, and develops an attribute recovery module to revert the injected node embedding from the graph embedding space back to the original feature space, enabling the adversary to manipulate node perturbation effectively.
- 13
- 3D model retrieval based on multi-view attentional convolutional neural network13
A discriminative Multi-View Attentional Convolutional Neural Network, dubbed as MVA-CNN, which takes the multiple views of an shape as input and output the object category, which can discover the context among multiple views in both the visual and spatial domain.
- Effective 3D object detection based on detector and tracker13
This paper proposes a novel 3D object detection approach in real-world scenes combined RGB image based on MV-RED dataset and shows further major gains in accuracy when the training data from the target scenes is used, outperforming state-of-the-art approaches with far better efficiency.
- 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
- 12
- 12
- 12
- 12
- 12
- SVHAN: Sequential View Based Hierarchical Attention Network for 3D Shape Recognition12
This work proposes a novel sequential view based hierarchical attention network (SVHAN) for 3D shape recognition, and introduces a novel hierarchical feature aggregation module (HFAM) and a novel selective fusion module (SFM) for feature aggregation.
- 12
- 12
- Open-view human action recognition based on linear discriminant analysis12
This work introduces a novel multi-view surveillance action dataset and proposes a novel open-view action recognition method based on the linear discriminant analysis that can learn a common space for action samples under different view by using their category information, which can achieve a good result in open-View action recognition.
- Sparse Coding Induced Transfer learning for HEp-2 Cell Classification12
This work is the first to transfer the human-crafted visual feature, sensitive to the variation of appearance and shape during cell movement, to the high level representation which directly denotes the correlation of one sample and the bases in the learnt dictionary.
- 12
- 11
- 11
- 11
- Decomposed Prototype Learning for Few-Shot Scene Graph Generation11
A novel Decomposed Prototype Learning (DPL) model is proposed, which first construct a decomposable prototype space to capture diverse semantics and visual patterns of subjects and objects for predicates by decomposing them into multiple prototypes to generate query-adaptive predicate representation with more reliable semantics for each query sample.
- 11
- 11
- Self-supervised Image-based 3D Model Retrieval11
This article simultaneously optimize the adversarial adaptation for both domains based on visual features and the contrastive learning for unlabeled 3D model domain to help the feature extractor to learn discriminative feature representations.
- Semantic and Context Information Fusion Network for View-Based 3D Model Classification and Retrieval11
This paper proposes a novel Semantic and Context information Fusion Network (SCFN), which renders views from multiple perspectives of the 3D model and extracts the raw feature of the individual view by 2D convolutional neural networks (CNN), and designs the channel attention mechanism (CAM) to exploit the view-wise semantic information.
- A comparison of methods for 3D scene shape retrieval11
A comprehensive comparison of all the above methods and an additional new retrieval methods are performed by evaluating them on the two benchmarks to solicit state-of-the-art approaches.
- Hierarchical Graph Structure Learning for Multi-View 3D Model Retrieval11
An unsupervised hierarchical graph structure learning method for 3D model retrieval that can avoid the difficulty in definition and computation in the traditional high-order graph and can achieve competing performance against the most recent supervised & deep learning method.
- SMTPD: A New Benchmark for Temporal Prediction of Social Media Popularity10
It is verified that temporal alignment and early popularity play crucial roles in social media popularity prediction for not only deepening the understanding of temporal dynamics of popularity in social media but also offering a suggestion about developing more effective prediction models in this field.
- 10
- 10
- 10
- 10
- 10
- 10
- Multi-graph Convolutional Network for Unsupervised 3D Shape Retrieval10
A novel multi-graph network (MGN) is proposed for unsupervised 3D shape retrieval, which utilizes the correlations among modalities and structural similarity between two models to guide the shape representation learning process without category information.
- FRSFN: A semantic fusion network for practical fashion retrieval10
This work presents a practical fashion retrieval method which puts emphasis on specific items, and grasps the detailed characteristics and items of the clothing and keeps a satisfying overall similarity in shape and color.
- Multi-scale CNNs for 3D model retrieval10
A novel multi-layer deep network structure is proposed, which owns the local and global information of a 3D object, and is used to compute the similarity between two different 3D models to complete the retrieval problem.
- 3D object retrieval based on Spatial+LDA model10
This paper applies LDA model for visual topics extraction and utilized the topic distribution visual feature of image to handle 3D object retrieval problem and proposes a multi-topic model to improve retrieval performance.
- 9
- Privacy-preserving Multi-source Cross-domain Recommendation Based on Knowledge Graph9
This work constructs a knowledge graph comprehensively leveraging items from multiple auxiliary domains to avoid the leakage of user privacy when user information is transferred to multiple domains, and proposes a knowledge graph-based retrieval and fusion method to transfer knowledge derived from multiple auxiliary domains.
- 9
- 9
- 9
- 9
- 9
- Multi-guiding long short-term memory for video captioning9
A novel cell-wise ensemble, where the weight matrix of each cell of LSTM is extended to be a set of attribute-dependent and attention-dependent weight matrices, by which the guiders induce each cell optimization over time.
- Structured low-rank inverse-covariance estimation for visual sentiment distribution prediction9
The proposed model incorporates low-rank and inverse-covariance regularization terms into a unified framework to learn more robust feature representation and more reasonable prediction model simultaneously.
- Watching, Thinking, Reacting: A Human-Centered Framework for Movie Content Analysis9
The promising results of subjective and objective evaluation indicate that the proposed framework can make not only computers intelligently understand movie content, but also provide personalized service for movie highlights recommendation to effectively lead audiences to preview new movies in an individualized manner.
- 9
- 8
- 8
- 8
- 8
- Bots Shield Fake News: Adversarial Attack on User Engagement based Fake News Detection8
This study introduces an attack problem that perturbs user-news engagements by injecting bots to shield the targeted fake news from being detected by GNN-based fake news detection models, and proposes a black-box attack method named Query-enhanced Surrogate-based Attack under Assortativity Constraint (QSA-AC) to work for this attack problem.
- Improving Rumor Detection by Class-based Adversarial Domain Adaptation8
Experiments on three public datasets show that CADA can improve the detection performance of existing rumor detection models and achieve better results than state-of-the-art models.
- Instrumental Variable Learning for Chest X-ray Classification8
This paper focuses on the chest X-ray classification task and proposes an interpretable instrumental variable (IV) learning framework, to eliminate the spurious association and obtain accurate causal representation.
- Intra-Class Adaptive Augmentation With Neighbor Correction for Deep Metric Learning8
A novel intra-class adaptive augmentation (IAA) framework for deep metric learning is proposed and this method significantly improves and outperforms the state-of-the-art methods on retrieval performances by 3%-6%.
- 8
- Consistent Domain Structure Learning and Domain Alignment for 2D Image-Based 3D Objects Retrieval8
A domain alignment framework with consistent domain structure learning to reduce the large gap between 2D images and 3D objects and the experimental results demonstrate the proposed method outperforms the state-of-the-art methods.
- Adaptive Ideal Image Reconstruction for Bacteria Colony Detection8
This method focuses on reconstructing ideal bacteria colony image by eliminating the bias of illumination in the reconstructed ideal image bacteria regions correspond to high intensity parts and the backgrounds lie in relatively low intensity parts.
- 7
- 7
- 7
- Cross-Modal Coherence-Enhanced Feedback Prompting for News Captioning7
This work proposes the novel cross-modal coherence-enhanced feedback prompting method to clarify the crucial elements that align closely with the visual content for news captioning.
- 7
- A comprehensive survey on deep-learning-based visual captioning7
This paper presents a comprehensive survey of existing deep learning-based visual captioning methods, relying on the adopted mechanism and technique to narrow the semantic gap, and divides visual captioned methods into various groups.
- Multi-type decision fusion network for visual Q&A7
The novel multi-type decision fusion model, which can adaptively leverage object- and relation-level decision networks to infer the answer, is introduced.
- End-to-End Visual Domain Adaptation Network for Cross-Domain 3D CPS Data Retrieval7
An end-to-end domain adaptation framework for cross-domain 3D objects retrieval (C3DOR-Net), which learns a joint embedding space for3D objects from different domains in an end- to-end manner, and focuses on the unsupervised case when 3D Objects in the target domain are unlabeled.
- 3D models retrieval algorithm based on multimodal data7
A novel algorithm based on multimodal 3D model data to handle model retrieval problem is proposed, and a universal graph matching is employed to handle similarity measure in different modals respectively.
- Quality models for venue recommendation in location-based social network7
A quality model for venue recommendation is proposed by utilizing multimedia data to predict the interested level of each venue and the weighted bipartite matching algorithm is implemented to generate the venue multimedia topics by bridging the textual information and the visual information.
- TRECVID 2006 Rushes Exploitation By CAS MCG7
This work proposes a novel and interactive rushes video selection and editing method based on hierarchical browsing of key frames, where high level features of each key frame such as face, interview, person, crowd, building, outdoor, waterbody, and other information about redundancy and repetition are displayed at same time for helping editors to select what they really want.
- Rushes Exploitation 2006 By CAS MCG7
This work proposes a novel and interactive rushes video selection and editing method based on hierarchical browsing of key frames, where high level features of each key frame such as face, interview, person, crowd, building, outdoor, waterbody, and other information about redundancy and repetition are displayed at same time for helping editors to select what they really want.
- 6
- 6
- IremulbNet: Rethinking the inverted residual architecture for image recognition6
The core module of this model is to reconstruct an inverse residual structure, in which a special feature fusion method, multi-branch feature extraction, and depthwise separable convolution techniques are used to improve the classification accuracy.
- Enhancing ENSO predictions with self-attention ConvLSTM and temporal embeddings6
A Two-Stage SpatioTemporal (TSST) autoregressive model that couples the meteorological factor prediction with ENSO indicator prediction and outperforms the state-of-the-art ENSO prediction models, effectively predicting ENSO up to 24 months and mitigating the spring predictability barrier.
- 6
- 6
- 6
- Cross-Domain 3D Model Retrieval Based On Contrastive Learning And Label Propagation6
This work proposes the consistent decision for pseudo labels of 3D models based on both the trained image classifier and label propagation and utilizes contrastive learning for fine-grained category-level alignment across domains.
- 6
- Low-rank regularized multi-view inverse-covariance estimation for visual sentiment distribution prediction6
This paper proposes a novel supervised visual sentiment distribution prediction model, termed as low-rank regularized multi-view inverse-covariance estimation, in which feature embedding and distribution prediction are jointly performed and develops an alternative heuristic optimization algorithm to solve the objective function of the proposed model.
- Special Issue on Biomedical Big Data: Understanding, Learning and Applications6
The papers in this special issue focus on biomedical Big Data, where the main challenge and bottleneck for the related research is the conversion of “biomedical big data” into interpretable information and hence discoveries.
- Multi-view feature extraction based on slow feature analysis6
This paper built the input data in order to guarantee it suitable for SFA mode according to structure information of3D model and employed graph matching to handle similarity measure problem between two different 3D model.
- 6
- 6
- 6
- 6
- T2IW: Joint Text to Image & Watermark Generation5
This T2IW framework is employed to ensure minimal damage to image quality when generating a compound image by forcing the semantic feature and the watermark signal to be compatible in pixels and strengthens the watermark robustness of this approach by subjecting the compound image to various post-processing attacks.
- 5
- 5
- Rule-Driven News Captioning5
This paper proposes the rule-driven news captioning method, which can generate image descriptions following designated rule signal and injects this semantic rule into the large-scale pre-trained model, BART, with the prefix-tuning strategy, where multiple encoder layers are embedded with news-aware semantic rule.
- SMPC: boosting social media popularity prediction with caption5
This paper proposes social media popularity prediction with caption (SMPC), a novel architecture that integrates the caption as the global representation into the existing multi-model-feature-based popularity prediction method.
- Multi-stage reasoning on introspecting and revising bias for visual question answering5
A bias discrimination module is proposed to distinguish good and bad bias indiscriminately in existing debiasing methods and helps with feature updating for both image regions and question words, thereby facilitating more accurate and comprehensive reasoning about both the image content and question.
- 5
- 5
- Semantically guided projection for zero-shot 3D model classification and retrieval5
This work proposes the semantically guided projection method to classify and retrieve unseen3D models by exploring the semantic relationship between seen and unseen 3D models, and learns bidirectional projections from visual features to semantics and from semantics to visual features, which can eliminate the gap between Seen and unseen domains.
- 5
- 5
- Unsupervised Feature Learning With Graph Embedding for View-Based 3D Model Retrieval5
The proposed method utilizes both similarity and dissimilarity information to better leverage the unsupervised problem and estimates the labels which are further used for metric learning.
- Multi-Camera Action Dataset (MCAD): A Dataset for Studying Non-overlapped Cross-Camera Action Recognition.5
A new dataset, namely Multi-Camera Action Dataset (MCAD), is introduced, which is designed to evaluate the open view classification problem under the surveillance environment and designed to design a standard evaluation protocol to enable comparison against existing works.
- A Hierarchical Framework for Mitosis Detection in Time-Lapse Phase Contrast Microscopy Image Sequences of Stem Cell Populations5
The need for extended-time observation and the proliferation of high-throughput imaging have made automated image analysis mandatory and phase contrast microscopy is a superior imaging modality because it enables continuous monitoring of live cells without requiring destructive methods of cell manipulation, such as cell lysis and staining.
- 5
- 4
- 4
- 4
- 4
- Cross-Modal Contrastive Learning with a Style-Mixed Bridge for Single Image 3D Shape Retrieval4
A style-mixed cross-modal contrastive learning method to mix the styles of images and rendered shape views to form an intermediate style and inject it into image contents in order to alleviate the modality gap.
- 4
- Social bot detection on Twitter: robustness evaluation and improvement4
A robustness evaluation method: Attribute Random Iteration-Fast Gradient Sign Method (ARI-FGSM) is proposed and uses a simplified adversarial training to improve the robustness of social bot detection.
- Model Can Be Subtle: Two Important Mechanisms for Social Media Popularity Prediction4
This article divides social network users into two categories and finds a dilemma in existing models and proposes the knowledge embedding mechanism to revise the popularity scores in existing models by fusing the statistical frequency over multi-type data.
- 4
- Multi-Modal Meta-Transfer Fusion Network for Few-Shot 3D Model Classification4
A multi-modal meta-transfer fusion network that can adaptively integrate the multi-modal information based on their significance to the 3D model for a more robust 3D representation is proposed.
- Progressive Positive Association Framework for Image and Text Retrieval4
A novel Progressive Positive Association Framework is proposed, which models association uncertainty as a normal distribution and progressively mines direct and potential positive associations according to the characteristics of the association distribution and applies KL loss constraint on cross-modal association distribution in order to enhance local semantic alignment.
- 4
- Prototype-based semantic consistency learning for unsupervised 2D image-based 3D shape retrieval4
The prototype-based semantic consistency (PSC) learning method is proposed, exploring semantic knowledge in both prototype-prototype and prototype-instance relationships in the probability space rather than the embedding space to preserve the structure of semantic information.
- External Knowledge Dynamic Modeling for Image-text Retrieval4
An External Knowledge Dynamic Modeling~(EKDM) architecture based on the filtering mechanism, which dynamically explores different knowledge towards varied image-text pairs and integrates the variable spatial locations between objects for association exploration.
- 4
- Evolution of ICTs-empowered-identification: A general re-ranking method for person re-identification4
A Feature Relation Map to mine the latent relation between the k -neighbors through convolution neural network is introduced, and a metric learning based Similarity Evaluation (SE) model is proposed to obtain the re-ranking distance from the FRM by incorporating the dilated convolution and normal convolution kernels.
- 4
- Mitosis detection in biomedical images4
This chapter introduces the mitosis process during the cell life cycle, and gives a detailed description of theMitosis detection problem, and introduces some advanced research in mitosis detection areas in recent years.
- PGNet: Progressive Feature Guide Learning Network for Three-dimensional Shape Recognition4
This article proposes a novel progressive feature guide learning network (PGNet) that focuses on the correlation among multiple views and integrates multiple modalities for 3D shape recognition and proposes two information fusion schemes from visual and feature aspects.
- 4
- 4
- 3
- Counterfactual GAN for debiased text-to-image synthesis3
A new debiased text-to-image synthesis method with a counterfactual learning scheme that conceptualizes the process as calculating the Natural Direct Effect on the generated visual features and use the Total Indirect Effect to produce the final image for debiased text-to-image generation.
- 3
- 3
- 3
- Universal unsupervised cross-domain 3D shape retrieval3
A novel universal unsupervised domain adaptation network that introduces two kinds of discriminators, source-domain discriminator, and cycle cross-domain discriminator to reduce source domain bias and constructed two novel datasets.
- 3
- Closed-loop reasoning with graph-aware dense interaction for visual dialog3
This work proposes closed-loop reasoning with graph-aware dense interaction, aiming to discover cues through the dynamic structure of graph and leverage it to benefit dialog and image features.
- 3
- 3
- Identification of novel hub genes and lncRNAs related to the prognosis and progression of pancreatic cancer by microarray and integrated bioinformatics analysis3
DEGs and DElncRNAs were identified between PC tumors and matched AN tissues, and these transcripts were connected with malignant phenotypes in PC through different BPs and signaling pathways.
- Multi-View Tree Structure Learning for 3D Model Retrieval and Classification in Smart City3
This paper proposes the Multi-View Tree Structure (MVTS) learning for 3D model retrieval and recognition, which jointly realizes the feature learning, view-wise contextual information and tree spatial information encoding and view importance estimating, which enhances the discrimination of the3D model representation.
- Sequential Estimation of Frequency and Direction-of-Arrival Based on the Relaxed Coprime Array for Multiple Targets3
A relaxed coprime array based two-stage estimator, which can sequentially detect the frequencies and direction-of-arrival for multiple targets, guided by the closed-form robust algorithm yields a higher array sparsity compared with existing sparse arrays.
- Multi-View Graph Matching for 3D Model Retrieval3
A multi-view graph matching method (MVGM) for 3D model retrieval that can avoid the difficulty in the definition and computation of the traditional high-order graph and does not require a large-scale 3D dataset for model learning.
- Domain-Specific Alignment Network for Multi-Domain Image-Based 3D Object Retrieval3
The proposed method is the first unsupervised work to align multiple-domain 2D images with 3D objects in an end-to-end manner and the extensive experimental results demonstrate the superiority of the proposed method.
- View-based 3D model retrieval via supervised multi-view feature learning3
An effective view-based 3D model retrieval method via supervised multi-view feature learning (SMFL) via Singular Value Decomposition (SVD) algorithm, which helps to calculate the similarities between two different 3D models.
- View-Based 3D Model Retrieval via Multi-graph Matching3
A composition-based multi-graph matching method that seeks the optimal intermediate graph by diverse graph matching orders, according to the consistency of global matching, using the classic optimization method.
- MNetDec: A Flexible Algorithm for Metabolic Network Decomposition3
A novel network decomposition algorithm, called MNetDec, is proposed by integrating the dendrogram and modularity features of a metabolic network and shows that M netDec can generate biological reasonable modules with high modularity.
- Spatial Context Constrained Characteristic View Extraction for 3D Model Retrieval3
A view-based 3D model retrieval method based on spatial context constrained characteristic view extraction method that applies Bayesian model to compute the similarity between query model and candidate3D model to find the best matching 3D models.
- 2
- 2
- 2
- 2
- 2
- View sequence prediction GAN: unsupervised representation learning for 3D shapes by decomposing view content and viewpoint variance2
An unsupervised 3D shape representation learning method called View Sequence Prediction GAN (VSP-GAN), which decomposes view content and viewpoint variance, and can learn more discriminative features than the state-of-the-art methods.
- 2
- 2
- Reinforcement Learning Based Multi-modal Feature Fusion Network for Novel Class Discovery2
A Reinforcement Learning framework is employed to simulate the cognitive processes of humans for effectively addressing novel class discovery in the Open-set domain and enables effective control over the extent of exploration learning, ensuring the accuracy of learning in unknown data categories.
- Temporal Action Localization with Multi-temporal Scales2
Detailed experiments show that the proposed method can outperform state-of-the-art approaches on the THUMOS14 dataset and achieves comparable performance on the ActivityNet1.3 dataset.
- 2
- An End-to-End Perceptual Quality Assessment Method via Score Distribution Prediction2
This paper proposes to predict the distribution of opinion scores via an end-to-end convolutional neural network based on a pre-trained ResNet with 50 layers and a novel Statistical Region-of-Interest Pooling layer is introduced for lower model complexity, which enables effective training with few datum.
- Tianjin University and National University of Singapore at TRECVID 2017: Video to Text Description.2
This framework builds on the LSTM unit for sentence generation with one attention layer for sentence generation with one attention layer and is optimized with the combination of heterogeneous pair-wise examples.
- Function Annotation of Proteins in Eriocheir sinensis Based on the Protein-Protein Interaction Network2
The analysis of ribosome module indicates that the annotation of proteins and modules provides important references to be studied in the further in vivo experiment as well as the biological capability of the E. sinensis PIN.
- 2
- Free-VTON: Cost-free acceleration and quality enhancement for diffusion-based virtual try-on1
This work introduces an Adaptive Caching acceleration strategy that adaptively caches and reuses features according to the similarity between features in adjacent diffusion steps, which accelerates the inference process without affecting the quality of try on.
- When Headlines Meet Minds: Empowering News Recommendations with Social Simulator1
Inspired by social constructionism, Social Review-aware Recommendation (SRec) is proposed, a novel framework that integrates both objective content and the social review that integrates both objective content and the social review.
- 1
- 1
- 1
- Cross-Domain Adversarial Learning for Sea Surface Temperature Super-Resolution1
A novel cross-domain adversarial learning model is proposed that utilizes both microwave and infrared SST data for super-resolution task and a random adversarial classifier is proposed to dynamically alter adversarial samples during training, enabling the model to focus on global properties and intrinsic patterns rather than specific regional characteristics, thus achieving more consistent and generalized performance across different areas.
- 1
- 1
- 1
- 1
- 1
- 3D model retrieval via single image based on feature mapping1
A novel 3D model retrieval algorithm by cross-modal feature mapping (CMFM), which utilize one single image as query information to address 3D models retrieval problem, and is the first method to handle3D model retreival based on one single 2D image.
- Geo-location driven image tagging via cross-domain learning1
A novel cross-domain learning method for automatic image annotation with geo-location information is proposed and the topic model-based method for popular concept extraction to adaptively construct cross- domain datasets is proposed.
- An Effective Tracking System for Multiple Object Tracking in Occlusion Scenes1
An effective multi-object tracking system which can handle the partial occlusion in the tracking process and employs the part-based model to localize the person and body parts in every frame is proposed.
- 1
- TJUT-TJU@TRECVID 2011: Surveillance Event Detection1
A spatio-temporal interest point detector is adopted, which explicitly encodes appearance features together with motion information, to extract robust point features in a sliding window and the bag-of-word (BoW) approach is employed.
- 1
- News Anchorperson Detection Algorithm Based on Spatio-Temporal Slice1
The innovation of the algorithm is that it can fulfill the adaptive detection of anchorperson by analyzing the structure of news video without depending on the special models or prior knowledge.
- –
- –
- –
- –
- SGoT-R1: Social Graph of Thought Reasoning-Enhanced Multimodal Large Language Model for Harmful Meme Detection–
A Social Graph of Thought Reasoning Enhancement (SGoTRE) framework for harmful meme detection is proposed, internalizing multi-opinion reasoning logic into a single multimodal large model SGoT-R1 to achieve efficient and interpretable harmful meme detection.
- Cognitive Distillation for Information Forensics: Towards Improved Hateful Meme Detection–
A Cognition-driven Information Forensics (CIF) framework for hateful meme detection is proposed, which integrates the collective cognition of diverse users into the entire process of information forensics and hateful content detection.
- –
- –
- How to Understand Named Entities: Using Commonsense for News Captioning–
Commonsense knowledge is exploited to understand named entities for news captioning by correlating the news content with commonsense in the wild, which helps an agent to distinguish semantically similar named entities and describe named entities using words outside of training corpora.
- Domain Adaptation from Generated Multi-Weather Images for Unsupervised Maritime Object Classification–
A novel domain adaptation approach is proposed that leverages AIMO (source domain) to address the problem of limited labeled data, unbalanced distribution and domain shift in RMO (target domain), enhance the generalization of source features with the Vision-Language Models such as CLIP, and propose a difficulty score for curriculum learning to optimize training process.
- –
- –
- –
- –
- Causality is all you need–
The Causal Graph Routing framework is proposed, an integrated causal scheme relying entirely on the intervention mechanisms to reveal the cause-effect forces hidden in data, and can surpass the current state-of-the-art methods on both Visual Question Answer and Long Document Classification tasks.
- –
- –
- –
- Multi-layers CNNs for 3D Model Retrieval–
A multi-layers CNNs (MLCNN) structure for 3D model representation is proposed and the Euclidean metric is leveraged to compute the similarity between two different 3D models to complete the retrieval problem.
- –
- Automatic Cell Cycle Localization Using Latent-Dynamic Conditional Random Fields–
This paper proposes an automatic cell cycle localization method based on the Latent-Dynamic Conditional Random Fields model, which can simultaneously model the visual dynamics within one stage and visual transition between adjacent stages in one mitosis sequence.
- TJU-TJUT@TRECVID 2015: Surveillance Event Detection.–
An event detection system evaluated in 2015 TRECVID-SED and an interactive visual analytics system, which can enable effective rank detection results with score relations and utilize user feedbacks to improve surveillance event detection are presented.
- Implicit Generalized Predictive Control Based on Vinyl Acetate Polymerization Rate–
An on-line estimation model of VAC polymerization rate was proposed in this paper based on recursive least square method, to which the implicit generalized predictive control (GPC) algorithm was applied.
- –
- –
- Multi-modal Interview Concept Detection for Rushes Exploitation–
Face detection and audio classification are implemented to detect "face" and "speech" concepts for each shot and by integrating audiovisual information, "interview" concept is finally detected.
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 2026-10-12. 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.