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
- Suggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works302 from public data
- The application of two-level attention models in deep convolutional neural network for fine-grained image classification829
This paper proposes to apply visual attention to fine-grained classification task using deep neural network and achieves the best accuracy under the weakest supervision condition, and is competitive against other methods that rely on additional annotations.
- Object-Part Attention Model for Fine-Grained Image Classification393
The object-part attention model (OPAM) for weakly supervised fine-grained image classification is proposed, which avoids the heavy labor consumption of labeling and compares with more than ten state-of-the-art methods on four widely-used datasets.
- An Overview of Cross-Media Retrieval: Concepts, Methodologies, Benchmarks, and Challenges341
An overview of cross-media retrieval is given, including the concepts, methodologies, major challenges, and open issues, as well as building up the benchmarks, including data sets and experimental results so that researchers can directly adopt the benchmarks to promptly evaluate their proposed methods.
- 294
- CM-GANs290
The proposed CM-GAN approach can use GANs to perform cross-modal common representation learning by which the heterogeneous data can be effectively correlated.
- Error-Driven Incremental Learning in Deep Convolutional Neural Network for Large-Scale Image Classification259
A training algorithm is developed that grows a network not only incrementally but also hierarchically, and divided into component models that predict coarse-grained superclasses and those return final prediction within a superclass.
- Unsupervised Generative Adversarial Cross-Modal Hashing233
This paper proposes an Unsupervised Generative Adversarial Cross-modal Hashing approach (UGACH), which makes full use of GAN's ability for unsupervised representation learning to exploit the underlying manifold structure of cross- modal data.
- CCL: Cross-modal Correlation Learning With Multigrained Fusion by Hierarchical Network227
This paper proposes a cross-modal correlation learning (CCL) approach with multigrained fusion by hierarchical network and compares with 13 state-of-the-art methods on 6 widely-used cross- modal datasets the experimental results show the CCL approach achieves the best performance.
- Cross-media shared representation by hierarchical learning with multiple deep networks196
The cross-media multiple deep network (CMDN) is proposed to exploit the complex cross- media correlation by hierarchical learning and achieves better performance comparing with several state-of-the-art methods on 3 extensively used cross- Media datasets.
- Object-Aware Aggregation With Bidirectional Temporal Graph for Video Captioning195
This paper proposes a new video captioning approach based on object-aware aggregation with bidirectional temporal graph (OA-BTG), which captures detailed temporal dynamics for salient objects in video, and learns discriminative spatio-temporal representations by performing object- aware local feature aggregation on detected object regions.
- Modality-Specific Cross-Modal Similarity Measurement With Recurrent Attention Network156
Modality-specific cross-modal similarity measurement approach is proposed by constructing the independent semantic space for each modality, which adopts an end-to-end framework to directly generate the modality- specific cross- modal similarity without explicit common representation.
- SSDH: Semi-Supervised Deep Hashing for Large Scale Image Retrieval156
This paper proposes a semi-supervised loss to jointly minimize the empirical error on labeled data, as well as the embedding error on both labeled and unlabeled data, which can preserve the semantic similarity and capture the meaningful neighbors on the underlying data structures for effective hashing.
- 153
- SCH-GAN: Semi-Supervised Cross-Modal Hashing by Generative Adversarial Network133
A novel semi-supervised cross-modal hashing approach by generative adversarial network (SCH-GAN) is proposed, in which the generative model tries to select margin examples of one modality from unlabeled data when given a query of another modality.
- SIM-Trans: Structure Information Modeling Transformer for Fine-grained Visual Categorization132
The proposed SIM-Trans achieves state-of-the-art performance on fine-grained visual categorization benchmarks and is light-weighted and can be plugged into any transformer network and trained end-to-end easily.
- 131
- Heterogeneous Metric Learning with Joint Graph Regularization for Cross-Media Retrieval129
A joint graph regularized heterogeneous metric learning (JGRHML) algorithm, which integrates the structure of different media into a joint graphRegularization, which is effective to explore the semantic relationship hidden across different modalities.
- MHTN: Modal-Adversarial Hybrid Transfer Network for Cross-Modal Retrieval128
A modal-adversarial hybrid transfer network (MHTN) is proposed, which aims to realize knowledge transfer from a single-modal source domain to a cross- modal target domain and learn cross-Modal common representation.
- Multi-Behavior Recommendation with Cascading Graph Convolution Networks127
A novel multi-behavior recommendation model with cascading graph convolution networks (named MB-CGCN), in which the embeddings learned from one behavior are used as the input features for the next behavior's embedding learning after a feature transformation operation.
- Two-Stream Collaborative Learning With Spatial-Temporal Attention for Video Classification127
The two-stream collaborative learning with spatial-temporal attention (TCLSTA) approach achieves the best performance compared with more than 10 state-of-the-art methods and adaptively learns the fusion weights of static and motion streams, thus exploiting the strong complementarity betweenstatic and motion information to improve video classification.
- 120
- 106
- 105
- Cross-media analysis and reasoning: advances and directions104
By presenting approaches, advances, and future directions in cross-media analysis and reasoning, the goal is to draw more attention to the state-of-the-art advances in the field, and to provide technical insights by discussing the challenges and research directions in these areas.
- 103
- 103
- PosterLayout: A New Benchmark and Approach for Content-Aware Visual-Textual Presentation Layout100
This work proposes design sequence formation (DSF) that reorganizes elements in layouts to imitate the design processes of human designers, and a novel CNN-LSTM-based conditional generative adversarial network (GAN) is presented to generate proper layouts.
- Weakly Supervised Learning of Part Selection Model with Spatial Constraints for Fine-Grained Image Classification100
This paper proposes a weakly supervised part selection method with spatial constraints for fine-grained image classification, which is free of using any bounding box or part annotations, and achieves a notable improvement on fine-grade image classification.
- 91
- 88
- Fast Fine-Grained Image Classification via Weakly Supervised Discriminative Localization88
A weakly supervised discriminative localization approach (WSDL) for fast fine-grained image classification to address the two limitations at the same time, and its main advantages are multi-level attention guided localization learning.
- Disentangled Graph Neural Networks for Session-Based Recommendation84
A novel method called Disentangled Graph Neural Network (Disen-GNN) is proposed to capture the session purpose with the consideration of factor-level attention on each item to take user intents at the factor level into account to infer the user purpose in a session.
- Which and How Many Regions to Gaze: Focus Discriminative Regions for Fine-Grained Visual Categorization80
Compared with state-of-the-art methods on two widely-used fine-grained visual categorization datasets, the M2DRL approach achieves the best categorization accuracy.
- A New Benchmark and Approach for Fine-grained Cross-media Retrieval77
A uniform deep model, namely FGCrossNet, which simultaneously learns 4 types of media without discriminative treatments is proposed, which is the first benchmark with 4 media types for fine-grained cross-media retrieval.
- DyFo: A Training-Free Dynamic Focus Visual Search for Enhancing LMMs in Fine-Grained Visual Understanding74
DyFo (Dynamic Focus), a training-free dynamic focusing visual search method that enhances fine-grained visual understanding in large multi-modal models (LMMs) and reduces hallucination issues in LMMs, achieving superior performance across both fixed and dynamic resolution models.
- Toward Video Anomaly Retrieval From Video Anomaly Detection: New Benchmarks and Model74
A novel task called Video Anomaly Retrieval (VAR), which aims to pragmatically retrieve relevant anomalous videos by cross-modalities, e.g., language descriptions and synchronous audios, and introduces an efficient pretext task to enhance semantic associations between video-text fine-grained representations.
- 73
- 73
- Fine-Grained Visual-Textual Representation Learning71
A fine-grained visual-textual representation learning (VTRL) approach, which achieves the best categorization accuracy compared with the state-of-the-art methods.
- Visual-textual Attention Driven Fine-grained Representation Learning.71
A fine-grained visual-textual representation learning (VTRL) approach, which achieves the best categorization accuracy compared with the state-of-the-art methods.
- 69
- 67
- IRC-GAN: Introspective Recurrent Convolutional GAN for Text-to-video Generation66
A novel Introspective Recurrent Convolutional GAN approach, where LSTM cells are integrated with 2D transconvolutional layers to generate plausible videos from given text and proposes mutual-information introspection to semantically align the generated video to text.
- Better and Faster: Knowledge Transfer from Multiple Self-supervised Learning Tasks via Graph Distillation for Video Classification65
Experimental results on 3 video datasets validate that the proposed graph-based distillation framework not only helps learn better video representation but also compress model for faster inference.
- Efficient Adaptive Human-Object Interaction Detection with Concept-guided Memory63
An efficient Adaptive HOI Detector with Concept-guided Memory (ADA-CM), Inspired by the powerful generalization ability of the large Vision-Language Models (VLM) on classification and retrieval tasks, which achieves competitive results with state-of-the-art on the HICO-DET and V-COCO datasets with much less training time.
- 62
- FineParser: A Fine-Grained Spatio-Temporal Action Parser for Human-Centric Action Quality Assessment60
It is argued that a fine-grained understanding of actions requires the model to perceive and parse actions in both time and space, which is also the key to the credibility and inter-pretability of the AQA technique.
- Query-Adaptive Image Retrieval by Deep-Weighted Hashing58
The query-adaptive deep weighted hashing approach is proposed, which can perform fine-grained ranking for different queries by weighted Hamming distance and outperforms eight state-of-the-art hashing methods.
- 58
- FinePOSE: Fine-Grained Prompt-Driven 3D Human Pose Estimation via Diffusion Models57
A new Fine-Grained Prompt-Driven Denoiser based on a diffusion model for 3D HPE, named FinePOSE, which outperforms state-of-the-art methods on public single-human pose estimation datasets and extends to multi-human pose estimation.
- MKVSE: Multimodal Knowledge Enhanced Visual-semantic Embedding for Image-text Retrieval57
A Multimodal Knowledge enhanced Visual-Semantic Embedding (MKVSE) approach building a multi-modal knowledge graph to explicitly represent the implicit multimodal knowledge relations and injecting it to visual-semantic embedding for image-text retrieval task is proposed.
- 56
- 56
- Cross-modal Common Representation Learning by Hybrid Transfer Network56
This paper proposes Cross-modal Hybrid Transfer Network (CHTN) with two subnetworks: Modal-sharing transfer subnetwork utilizes the modality in both source and target domains as a bridge, for transferring knowledge to both two modalities simultaneously.
- Visual Data Synthesis via GAN for Zero-Shot Video Classification55
A visual data synthesis framework via GAN is proposed to address information degradation issue, which captures seen-to-unseen correlation in matched and mismatched visual-semantic pairs by mutual information, providing the zero-shot synthesis procedure with robust guidance signals.
- Adaptive Trajectory Tracking Control of a Parallel Ankle Rehabilitation Robot With Joint-Space Force Distribution54
An adaptive trajectory tracking control strategy implemented on a parallel ankle rehabilitation robot with joint-space force distribution is proposed, demonstrating its potential in assisting ankle therapy.
- Deep Reinforcement Learning for Image Hashing54
This work proposes a new Deep Reinforcement Learning approach for Image Hashing (DRLIH), which models the hashing learning problem as a sequential decision process, which learns each hashing function by correcting the errors imposed by previous ones and promotes retrieval accuracy.
- 53
- Deep Cross-Media Knowledge Transfer50
Deep cross-media knowledge transfer (DCKT) approach is proposed, which transfers knowledge from a large-scale cross- media dataset to promote the model training on another small-scaleCross-media dataset to enhance the robustness of model training.
- Attribute-Aware Deep Hashing With Self-Consistency for Large-Scale Fine-Grained Image Retrieval47
Qualitative results demonstrate that not only the obtained hash codes can strongly correspond to certain kinds of crucial properties of fine-grained objects, but also the self-consistency designs can effectively overcome simplicity bias in fine- grained hashing.
- 47
- Multi-Scale Correlation for Sequential Cross-modal Hashing Learning44
The proposed Multi-scale Correlation Sequential Cross-modal Hashing Learning (MCSCH) approach proposes a multi-scale correlation mining strategy during the multi- scale feature guided sequential hashing learning, which can simultaneously mine the correlation among the features of different modalities and scales.
- Color-based clustering for text detection and extraction in image44
This paper proposes a new approach for the text detection and extraction in image that can adaptively select the best color plane according to the text contrast difference among color planes for binarization.
- The earth mover's distance as a semantic measure for document similarity43
A novel similarity measure based on the earth mover's distance (EMD) is proposed, which is employed to calculate the document similarity with a many-to-many matching between words.
- MV-Diffusion: Motion-aware Video Diffusion Model42
A Motion-aware Video Diffusion Model (MV-Diffusion) for enhancing the temporal consistency of generated videos using autoregressive diffusion models by leveraging rich temporal trajectory information in a global context and explicitly modeling local motion trends.
- Cross-modal Bidirectional Translation via Reinforcement Learning41
This work proposes Cross-modal Bidirectional Translation (CBT) approach, which attempts to treat images as a special kind of language to provide visual descriptions, so that translation can be conduct between bilingual pair of image and text to effectively explore cross- modal correlation.
- 40
- Attribute hierarchy based multi-task learning for fine-grained image classification40
Compared with more than 10 state-of-the-art methods on two widely-used CUB-200-2011 and Cars-196 datasets, the AHMTL approach achieves the best classification performance.
- Text-to-image Synthesis via Symmetrical Distillation Networks40
The excellent capability of generic discriminative models is exploited, which can guide the training process of a new generative model on multiple levels to bridge the two gaps, to propose Symmetrical Distillation Networks (SDN), composed of a source discrim inative model as "teacher" and a target generative models as "student".
- 39
- 39
- Analyzing and Boosting the Power of Fine-Grained Visual Recognition for Multi-modal Large Language Models38
This study presents Finedefics, an MLLM that enhances the model's FGVR capability by incorporating informative attribute descriptions of objects into the training phase, and employs contrastive learning on object-attribute pairs and attribute-category pairs.
- 38
- 38
- Effective Heterogeneous Similarity Measure with Nearest Neighbors for Cross-Media Retrieval38
A novel heterogeneous similarity measure with nearest neighbors (HSNN) is proposed which could compute the similarity between media objects with different media types and shows to outperform image retrieval systems on a unimedia retrieval task.
- Cross-media Multi-level Alignment with Relation Attention Network37
This work proposes Cross-media Relation Attention Network (CRAN) with multi-level alignment to explore global, local and relation alignments across different media types, which can mutually boost to learn more precise cross-media correlation.
- 36
- EMD-Based Video Clip Retrieval by Many-to-Many Matching35
Experimental results show that the proposed approach for video clip retrieval based on Earth Mover’s Distance is better than some existing methods in term of ranking capability.
- 34
- Scanning Only Once: An End-to-end Framework for Fast Temporal Grounding in Long Videos34
This work proposes an end-to-end framework for fast temporal grounding, which is able to model an hours-long video with one-time network execution, and significantly outperforms state-of-the-arts, and achieves 14.6× / 102.8× higher efficiency respectively.
- 34
- Image categorization via robust pLSA34
This paper presents a novel method to give a good initial estimate of the probabilistic latent semantic analysis (pLSA) model using rival penalized competitive learning (RPCL), since the expectation maximization (EM) algorithm used to train the model is sensitive to the initialization.
- Learning conditional photometric stereo with high-resolution features33
A self-learning conditional network with multi-scale features for photometric stereo, avoiding blurred reconstruction in high-frequency regions, and an improved gradient-motivated conditionally parameterized convolution (GM-CondConv) in the network.
- From comparing clusterings to combining clusterings32
This paper presents a fast simulated annealing framework for combining multiple clusterings based on some measures of agreement between partitions, which are originally used to compare two clusterings for the evaluation of a clustering algorithm.
- Dual Adversarial Networks for Zero-shot Cross-media Retrieval31
The DANZCR approach consists of two GANs in a dual structure for common representation generation and original representation reconstruction respectively, which capture the underlying data structures as well as strengthen relations between input data and semantic space to generalize across seen and unseen categories.
- 30
- 30
- 29
- Hierarchical Vision-Language Alignment for Video Captioning29
An Attention Guided Hierarchical Alignment (AGHA) approach for video captioning, where hierarchical vision-language alignments are extracted from a well-learned model that suits for multiple tasks related to vision and language, which are embedded into parallel encoder-decoder streams to provide multi-level semantic guidance and rich complementarities on description generation.
- Exemplar-Free Lifelong Person Re-identification via Prompt-Guided Adaptive Knowledge Consolidation28
This paper proposes a paradigm for exemplar-free LReID through model re-parameterization, and adopts a novel Prompt-guided Adaptive Exponential Moving Average (PAEMA) strategy to achieve dynamic knowledge consolidation.
- Generating Structured Pseudo Labels for Noise-resistant Zero-shot Video Sentence Localization28
A Structure-based Pseudo Label generation (SPL) is proposed, which first generate free-form interpretable pseudo queries before constructing query-dependent event proposals by modeling the event temporal structure.
- Fuzzy logic compliance adaptation for an assist-as-needed controller on the Gait Rehabilitation Exoskeleton (GAREX)28
It is demonstrated that FLCA controller could effectively distinguish the capability/effort levels of wearers and enable the exoskeleton to adapt the knee joint compliance accordingly and indicate the novel method of realizing assist-as-needed concept with the pneumatic muscle powered mechanisms.
- Fine-Grained Visual Prompt Learning of Vision-Language Models for Image Recognition27
Fine-grained visual prompt is introduced into the image encoder of the vision-language model for focusing on the target object and conducting information interaction within the object, which facilitates generating discriminative visual features for image recognition.
- Phrase-Level Temporal Relationship Mining for Temporal Sentence Localization27
A phrase-level Temporal Relationship Mining (TRM) framework employing the temporal relationship relevant to the phrase and the whole sentence to have a better understanding of each semantic entity in the sentence.
- L1-graph construction using structured sparsity27
A new graph is constructed, called the k -nearest neighbor ( k -NN) fused Lasso graph, which is different from the traditional L 1 -graph because of its successful incorporation of the structured sparsity into the graph construction process and its applicability to large complex datasets.
- Latent semantic learning with structured sparse representation for human action recognition26
This paper develops a graph-based spectral embedding approach to latent semantic learning, with the graph over mid-level features being constructed using sparse representation, and defines new L"1-norm hypergraph regularization to induce extra structured sparsity into sparse representation for graph construction.
- Cross-media retrieval by intra-media and inter-media correlation mining26
A novel cross-media similarity measure that considers both intra-media and inter-media correlation, which are ignored by existing works, and is able to propagate the correlation between heterogeneous modalities.
- Compositional Prompting for Anti-Forgetting in Domain Incremental Learning25
This paper proposes a unified and effective paradigm named Compositional Prompting (C-Prompt) to mitigate the critical forgetting challenge in DIL for image classification tasks and leverages a prompt-guided Batch-wise Exponential Moving Average (BEMA) strategy to adaptively consolidate learned knowledge without retaining any exemplars.
- SPIRIT: Style-guided Patch Interaction for Fashion Image Retrieval with Text Feedback25
This work proposes a Style-guided Patch InteRaction approach for fashion Image retrieval with Text feedback (SPIRIT), which focuses on the decisive influence of local details of fashion images on their style.
- 25
- Mining intricate temporal rules for recognizing complex activities of daily living under uncertainty25
A probabilistic semantic-based framework is presented that combines Markov logic network with 15 temporal and hierarchical relations to explicitly perform diverse inference tasks of daily living in a unified manner and is robust to the incomplete or incorrect observations of intervals.
- 25
- Clip-based similarity measure for hierarchical video retrieval25
This paper proposes a new approach and algorithm for the similarity measure of video clips, mainly based on two bipartite graph matching algorithms: maximum matching (MM) and optimal matching (OM).
- 24
- PosterO: Structuring Layout Trees to Enable Language Models in Generalized Content-Aware Layout Generation24
PosterO is a layout-centric approach that leverages layout knowledge implicit in large language models (LLMs) to create posters for omnifarious purposes, hence the name, and structures layouts from datasets as trees in SVG language by universal shape, design intent vectorization, and hierarchical node representation.
- LFR-GAN: Local Feature Refinement based Generative Adversarial Network for Text-to-Image Generation24
A Local Feature Refinement Based Generative Adversarial Network (LFR-GAN) is proposed, which first divides the text into some independent fine-grained attributes and generates an initial image, then refines the image details based on these attributes.
- Prototype-based classifier learning for long-tailed visual recognition24
This paper tackles the long-tailed visual recognition problem from the categorical prototype perspective by proposing a prototype-based classifier learning (PCL) method, and proves that the proposed PCL can outperform state-of-the-arts.
- 23
- Learn from Unlabeled Videos for Near-duplicate Video Retrieval23
Comprehensive experiments on two challenging near-duplicate video retrieval datasets, namely FIVR-200K and SVD, verify the effectiveness of the proposed VRL approach, which achieves the best performance of video retrieval on accuracy and efficiency.
- A framework of mining semantic-based probabilistic event relations for complex activity recognition23
A framework to learn an event knowledge base (EKB) of probabilistic interval-based event relations and use them to infer varied semantic-level queries about activity occurrences under uncertainty is presented.
- 23
- Latent Semantic Learning by Efficient Sparse Coding with Hypergraph Regularization23
Different from the traditional latent semantic analysis based on topic models, this sparse coding method with hypergraph regularization can exploit the manifold structure hidden among mid-level features for latent semantic learning, which results in compact but discriminative high- level features for action recognition.
- FashionERN: Enhance-and-Refine Network for Composed Fashion Image Retrieval22
A Triple-branch Modifier Enhancement model is introduced, which injects relevant information from the reference image and aligns the modified text modality with the target image modality, and a Dual-guided Vision Refinement model is proposed that retains critical visual information through text-guided refinement and self-guided refinement processes.
- 22
- 22
- Sequential Cross-Modal Hashing Learning via Multi-scale Correlation Mining22
The Multi-scale Correlation Sequential Cross-modal Hashing (MCSCH) approach, proposed to share the information from features of different scales through an RNN-based network and generate the hash codes sequentially, and reveals the finer underlying structure of multimedia data and can help to boost the hash function learning.
- Symmetric Graph Regularized Constraint Propagation22
It is clearly shown that pairwise constraint propagation is actually equivalent to solving a Lyapunov matrix equation, which is widely used in Control Theory as a standard continuous-time equation.
- 21
- StackDRL: Stacked Deep Reinforcement Learning for Fine-grained Visual Categorization21
A stacked deep reinforcement learning approach (StackDRL), which adopts a two-stage learning architecture, which is driven by the semantic reward function, and achieves the best categorization accuracy on CUB-200-2011 dataset.
- Unified Constraint Propagation on Multi-View Data21
This paper presents a unified framework for intra-view and inter-view constraint propagation on multi-view data and decomposes these two types of constraint propagation into semi-supervised learning subproblems so that they can be uniformly solved based on the traditional label propagation techniques.
- Image annotation by semantic sparse recoding of visual content21
A new semantic sparse recoding method to generate more descriptive and robust representation of visual content for image annotation by developing an efficient sparse coding algorithm and successfully generating a new visual BOW representation for image annotations.
- AdaOUBoost: adaptive over-sampling and under-sampling to boost the concept learning in large scale imbalanced data sets21
The AdaOUBoost is proposed, which aims to use fully the whole training data and improve the performance of the class-imbalance learning classifier, and can achieve the superior performance in large scale imbalanced data sets.
- 20
- Mitigate Catastrophic Remembering via Continual Knowledge Purification for Noisy Lifelong Person Re-Identification20
This paper initially investigates noisy LReID and proposes a Continual Knowledge Purification (CKP) method to address the catastrophic remembering of erroneous knowledge and catastrophic forgetting of correct knowledge simultaneously.
- 20
- 20
- 19
- Cross-modal deep metric learning with multi-task regularization19
This paper proposes Cross-modal Deep Metric Learning with Multi-task Regularization (CDMLMR), which integrates quadruplet ranking loss and semi-supervised contrastive loss for modeling cross-modAL semantic similarity in a unified multi-task learning architecture.
- 18
- Subject-specific compliance control of an upper-limb bilateral robotic system18
Findings indicate that human users can follow predefined training trajectories well under real-time compliance variation, with the maximum normalized root mean square error (NRMSE) value no greater than 1.44%.
- A New Retrieval Model Based on TextTiling for Document Similarity Search18
A new retrieval model based on TextTiling is proposed in the paper, which takes into account the subtopic structures of documents and outperforms other models, including the Cosine measure.
- 17
- DART: Dual-Modal Adaptive Online Prompting and Knowledge Retention for Test-Time Adaptation17
To increase the learning capacity, DART captures knowledge from each test sample by learning class-specific text prompts and instance-level image prompts, and utilizes dual-modal knowledge retention prompts to adaptively retain the acquired knowledge, thereby enhancing the predictions on subsequent test samples.
- CAT: a coarse-to-fine attention tree for semantic change detection17
This paper tackles the SCD and LCM tasks simultaneously by proposing a coarse-to-fine attention tree (CAT) model, which outperforms state-of-the-art models and equip a label distribution learning loss upon the model to address label ambiguity in SCD/LCM.
- Rapid characterization of anti-CRISPR proteins and optogenetically engineered variants using a versatile plasmid interference system17
A versatile plasmid interference with CRISPR interference (PICI) system in Escherichia coli is established for rapidly characterizing Acrs and developing Acr-based technologies, and two novel type II-A Acrs are discovered, which can inhibit the activity of SpyCas9 by affecting DNA recognition of Cas9.
- Semantic concept annotation based on audio PLSA model17
A new approach and algorithm for the semantic concept annotation based on audio PLSA (probabilistic latent semantic analysis) model, which combines the local features extracted by PLSA and global features in audio-clip to further improve the performance of concept annotation.
- ConMo: Controllable Motion Disentanglement and Recomposition for Zero-Shot Motion Transfer16
ConMo is introduced, a zero-shot framework that disentangle and recompose the motions of subjects and camera movements that unlocks a wide range of applications, including subject size and position editing, subject removal, semantic modifications, and camera motion simulation.
- 16
- 16
- 16
- DV-Net: Dual-view network for 3D reconstruction by fusing multiple sets of gated control point clouds16
Both the qualitative results and quantitative analysis demonstrate that this method not only captures the detailed geometric structures of 3D shapes for different object categories with complex topologies, but also achieves state-of-the-art performance.
- 16
- Cross-View Feature Learning for Scalable Social Image Analysis16
Through explicitly modeling the relevance between image content and tags, CVFL yields more promising results than existing methods in the experiments and can be readily applied to other large-scale multi-view tasks in unsupervised setting.
- Hot Event Detection and Summarization by Graph Modeling and Matching16
Based on the cluster size and globality of events, hot events can be automatically detected and selected as the summaries of news videos across TV stations of various channels and languages.
- STOP: Integrated Spatial-Temporal Dynamic Prompting for Video Understanding15
An integrated Spatial-TempOral dynamic Prompting (STOP) model which consists of two complementary modules, the intra-frame spatial prompting and inter-frame temporal prompting, which enables the model to prioritize key frames and enhances its capacity to understand temporal dependencies across sequences.
- InsVP: Efficient Instance Visual Prompting from Image Itself15
The instance image and feature prompt is introduced to extract both crucial and nuanced discriminative information from the original image itself and is overlaid onto the input image and fed into the model's intermediate layers to facilitate feature extraction.
- Continual Vision-Language Retrieval via Dynamic Knowledge Rectification15
A novel framework called Dynamic Knowledge Rectification (DKR) is proposed that simultaneously achieves incorrect knowledge filtering and rectification and introduces paired ground-truth labels to promote the acquisition of both old and new knowledge.
- Real20M: A Large-scale E-commerce Dataset for Cross-domain Retrieval15
The PKU Real20M dataset is proposed, a large-scale e-commerce dataset designed for cross-domain retrieval and the Query-driven Cross-Domain retrieval framework (QCD), which leverages user queries to facilitate efficient alignment between the product and micro-video domains.
- MB-HGCN: A Hierarchical Graph Convolutional Network for Multi-behavior Recommendation15
MB-HGCN is proposed, a novel multi-behavior recommendation model that uses a hierarchical graph convolutional network to learn user and item embeddings from coarse-grained on the global level to fine-graining on the behavior-specific level and adopts multi-task learning for optimization.
- 15
- Discriminative latent semantic feature learning for pedestrian detection15
A novel approach of learning discriminative features is proposed, addressing two main limitations of the methods in the literature by extending the learning process via a maximum margin criterion, in order to better encode discrim inative information directly in features.
- SCAP: Transductive Test-Time Adaptation via Supportive Clique-based Attribute Prompting14
A novel transductive TTA framework, Supportive Clique-based Attribute Prompting (SCAP), which effectively combines visual and textual information to enhance adaptation by generating fine-grained attribute prompts across test batches, significantly advancing VLM generalization under domain shifts.
- Semantic Graph Construction for Weakly-Supervised Image Parsing14
This work investigates weakly-supervised image parsing, proposes novel criteria by exploiting the weak supervision information carefully, and develops two graphs: L1 semantic graph and k-NN semantic graph, which perform significantly better than conventional graphs in image parsing.
- Advancing 3D Scene Understanding with MV-ScanQA Multi-View Reasoning Evaluation and TripAlign Pre-training Dataset13
This work introduces MV-ScanQA, a novel 3D question answering dataset where 68% of questions explicitly require integrating information from multiple views (compared to less than 7% in existing datasets), thereby rigorously testing multi-view compositional reasoning, and develops LEGO, a baseline method for the multi-view reasoning challenge in MV-ScanQA.
- EOGT: Video Anomaly Detection with Enhanced Object Information and Global Temporal Dependency13
A video anomaly detection framework with Enhanced Object Information and Global Temporal Dependencies (EOGT) and the main novelties are a Local Object Anomaly Stream (LOAS) and a Global Temporal Strengthening Stream (GTSS) with video-level temporal dependencies, which leverages video-level temporal dependencies to identify long-term and video-specific anomalies effectively.
- SIA-OVD: Shape-Invariant Adapter for Bridging the Image-Region Gap in Open-Vocabulary Detection13
A new Shape-Invariant Adapter named SIA-OVD is proposed to bridge the image-region gap in OVD and effectively improves the classification accuracy for regions by addressing the gap between images and regions caused by shape deformation.
- 13
- Graph-based multimodal semi-supervised image classification13
A graph-based multimodal semi-supervised image classification (GraMSIC) framework to handle an image classification task where training images come along with tags, but only a subset being labeled, and the goal is to predict the class label of test images without tags is proposed.
- Weakly-Supervised Image Parsing via Constructing Semantic Graphs and Hypergraphs13
This paper develops two novel graphs: Semantic graph, which is established upon a conventional graph by utilizing the proposed weakly-supervised criteria; 2) Semantic hypergraph, which explores both intra-image and inter-image high-order semantic relevance.
- 13
- 13
- Heterogeneous Graph Neural-Network-Based Scheduling Optimization for Multi-Product and Variable-Batch Production in Flexible Job Shops12
An intelligent scheduling approach based on a heterogeneity-enhanced graph neural network combined with deep reinforcement learning that achieves superior scheduling performance but also maintains an excellent balance between response efficiency and scheduling quality is presented.
- Open-Vocabulary Hoi Detection With Interaction-Aware Prompt and Concept Calibration12
An interaction-aware prompt generator that dynamically generates a compact set of prompts based on the input scene, enabling selective sharing among similar interactions and refine HOI concept representations through language modelguided calibration, which helps distinguish diverse HOI concepts by investigating visual similarities across categories.
- 12
- Only Learn One Sample12
Experimental results and analyses on fine-grained visual categorization benchmark demonstrate that the proposed OLOS approach can be applied on top of existing methods, and improves their categorization performance.
- Saliency-guided video classification via adaptively weighted learning12
Experimental results show that saliency-guided modeling and adaptively weighted learning can reinforce each other, and the states-of-the-art results are achieved.
- 11
- 11
- Exploiting distinctive topological constraint of local feature matching for logo image recognition11
A robust matching method is proposed to get the better matching results by exploiting the distinctive topological constraint together with the feature selection, which makes full use of the affine-invariant geometric relationship among adjacent local features for more accurate local feature matching.
- Adaptive Sampling with Optimal Cost for Class-Imbalance Learning11
The novelty of the proposed approach mainly lies in adaptively over-sampling the minority positive examples and under-sampled the majority negative examples, forming different sub-classifiers by different subsets of training data with the best cost ratio adaptively chosen, and combining these sub- classifiers according to their accuracy to create a strong classifier.
- 10
- 10
- Balancing Preservation and Modification: A Region and Semantic Aware Metric for Instruction-Based Image Editing10
A new metric called Balancing Preservation and Modification (BPM) is introduced, tailored for instruction-based image editing by explicitly disentangling the image into editing-relevant and irrelevant regions for specific consideration, and shows the highest alignment with human evaluation compared to existing metrics, indicating its efficacy.
- Firzen: Firing Strict Cold-Start Items with Frozen Heterogeneous and Homogeneous Graphs for Recommendation10
A unified framework incorporating multi-modal content of items and KGs to effectively solve both strict cold-start and warm-start recommendation termed Firzen is proposed, which extracts the user-item collaborative information over frozen heterogeneous graph (collaborative knowledge graph), and exploits the item-item semantic structures and user-user behavioral association over frozen homogeneous graphs.
- 10
- 10
- 10
- 10
- Group Cost-Sensitive Boosting for Multi-Resolution Pedestrian Detection10
A new multi-resolution detection approach based on a novel group cost-sensitive boosting algorithm, which extends the popular AdaBoost by exploring different costs for different resolution groups in the boosting process, and places more emphases on low resolution group in order to better handle detection of hard samples.
- 10
- 10
- 9
- 9
- ResVG: Enhancing Relation and Semantic Understanding in Multiple Instances for Visual Grounding9
The proposed ReSVG model significantly improves the model's ability to comprehend both object semantics and spatial relations, leading to enhanced performance in visual grounding tasks, particularly in scenarios with multiple-instance distractions.
- FineFMPL: Fine-grained Feature Mining Prompt Learning for Few-Shot Class Incremental Learning9
A Fine-grained Feature Mining Prompt Learning (FineFMPL) approach to adapt the VL model to FSCIL, which comprehensively learns and memorizes fine-grained discriminative information of emerging classes.
- 9
- Nonlinear color space coded by additive digital pulses9
A complete theory that encodes and decodes digital signals directly to a perceptually nonuniform color space of the commission internationale de l’eclairage is developed, featuring a strict bijection between the duty cycles of each binary pulse and color components of the mixed light and an analytic nonlinear gamut volume.
- Life-long Cross-media Correlation Learning9
This paper proposes Cross-media Life-long Learning (CmLL) approach, which can leverage the knowledge learned from the existing data, to obtain better correlation performance in new domain, and proposes both intra-domain distribution alignment as well as inter-domain knowledge distillation that can not only effectively preserve the correlation ability in old cross-media domains, but also improve the performance innew domain by transferring knowledge among different domains.
- Cross-media similarity metric learning with unified deep networks9
The proposed Unified Network for Cross-media Similarity Metric (UNCSM) approach unifies the representation learning and distance metric to preserve the relative similarity as well as embrace more complex similarity functions for further improving the cross-media retrieval accuracy.
- Identity-Preserving Text-to-Video Generation via Training-Free Prompt, Image, and Guidance Enhancement8
A Training-Free Prompt, Image, and Guidance Enhancement (TPIGE) framework that bridges the semantic gap between the video description and the reference image and design sampling guidance that enhance identity preservation and video quality, achieving performance gains at minimal cost is introduced.
- 8
- RelScene: A Benchmark and baseline for Spatial Relations in text-driven 3D Scene Generation8
This paper conducts a dataset and benchmark for assessing spatial relations in text-driven 3D scene generation, which contains a comprehensive collection of 3D scenes, including textual descriptions, annotating object spatial relations, and providing both template and free-form natural language descriptions.
- Smart Public Transportation Sensing: Enhancing Perception and Data Management for Efficient and Safety Operations8
Through operational testing in real environments, the proposed platform has proven highly efficient and effective in managing organizational operations, data assets, data life cycle, offline development, and backend administration over a large amount of various types of public transportation traffic data.
- Semantic association enhancement transformer with relative position for image captioning8
A Semantic Association Enhancement Transformer model is proposed that outperforms the state-of-the-art methods and is adaptively improved according to the constructed importance, especially the objects with weak dependence on appearance features, thereby enhancing their semantic association.
- 8
- 8
- Cross-Media Retrieval by Multimodal Representation Fusion with Deep Networks8
A deep multimodal learning method (DML) for cross-media shared representation learning is proposed, which can obtain the high-level semantic representation of single media.
- PKU-ICST at TRECVID 2014: Instance Search Task.8
This task mainly focused on exploring the effective feature representation, feature matching and re-ranking algorithm, and tried to use Deep Neural Networks (DNN) to improve the results.
- 8
- Audio similarity measure by graph modeling and matching8
The experimental result shows that the proposed approach for the similarity measure and ranking of audio clips by graph modeling and matching is better than some existing methods in terms of retrieval and ranking capabilities.
- Progressive Prototype Evolving for Dual-Forgetting Mitigation in Non-Exemplar Online Continual Learning7
The core of the PPE is to progressively learn class-specific prototypes during the online learning phase without reusing any previously seen data, and the progressive prototypes of the current learning stage are fed back to the model to mitigate intra-stage forgetting.
- 7
- Efficiency-optimized Video Diffusion Models7
Efficiency-optimized Video Diffusion Models can achieve a 10x speed-up while achieving comparable or even better results in the performance of fidelity compared with the state-of-the-art methods.
- 7
- A Semi-supervised Learning Algorithm on Gaussian Mixture with Automatic Model Selection7
The presented semi-supervised learning algorithm can automatically detect the number of Gaussian with a good parameter estimation, even when two or more actual Gaussians in the mixture are overlapped at a high degree.
- Comprehensive Visual Grounding for Video Description6
A comprehensive visual grounding network to improve video captioning, by explicitly linking the entities and actions to the visual clues across the video frames, which encourages the attention mechanism to focus on informative spatial areas across video frames.
- 6
- 6
- Zero-Shot Cross-Media Retrieval with External Knowledge6
This paper proposes a new approach to model both multimedia and external knowledge information, generated jointly by media features and category weight vectors which are learned by utilizing online encyclopedias.
- Object-Part Attention Driven Discriminative Localization for Fine-grained Image Classification.6
The object-part attention driven discriminative localization (OPADDL) approach for weakly supervised fine-grained image classification achieves the best performance compared with more than 10 state-of-the-art methods on 3 widely used datasets.
- Cross-Media Retrieval via Semantic Entity Projection6
A new approach to cross-media retrieval via semantic entity projection SEP, where an entity level with fine-grained semantics between low-level features and high-level concepts are constructed, so as to help bridge the semantic gap to a certain extent.
- 6
- 6
- OM-based video shot retrieval by one-to-one matching6
Experimental results indicate that the proposed approach for shot-based retrieval by optimal matching achieves better performance than other methods in terms of ranking and retrieval capability.
- 6
- RaT2IGen: Relation-aware Text-to-image Generation via Learnable Prompt5
RaT2IGen (Relation-aware Text-to-image Generation), which aims to improve the semantic consistency of the generation model of object relation with the proper spatial layout of the generated images with the proper spatial layout is proposed.
- 5
- 5
- 5
- Guest Editorial Introduction to the Special Section on Representation Learning for Visual Content Understanding5
This special section focuses on data-effective representation learning methods for visual content understanding using the ResNet method for image classification, the DeepFace method for face recognition, and the feature pyramid networks (FPNs) method for object detection.
- Recursive Pyramid Network with Joint Attention for Cross-Media Retrieval5
A joint attention model connecting different media types in each pyramid level is proposed, which mines the intra-media information and inter-media correlations to guide the learning of significance of each region, further improving the performance of correlation learning.
- Cross-media retrieval by exploiting fine-grained correlation at entity level5
The proposed approach first constructs an entity level with fine-grained semantics between low-level features and high-level concepts, and learns the distance-preserving entity projections (DPEP) and generates the unified descriptive representation of media content.
- Gaussian mixture learning via robust competitive agglomeration5
This work follows the idea of competitive agglomeration which is originally used for fuzzy clustering and proposes two robust algorithms for Gaussian mixture learning that can lead to automatic model selection on Gaussian mixtures and also make these algorithms less sensitive to initialization than the EM algorithm.
- 5
- Weakly and Single-Frame Supervised Temporal Sentence Grounding With Gaussian-Based Contrastive Proposal Learning4
This paper uses multiple learnable asymmetric Gaussian functions to generate both positive and negative proposals within the same video and proposes a controllable, easy-to-hard negative proposal mining strategy to collect negative samples within the same video, which enables CPL to distinguish highly confusing scenes.
- 4
- 4
- 4
- A temporal context model for boosting video annotation4
A temporal context model to mine the redundant information between shots to obtain the refined probability of a concept occurring in the shot, which is the leverage of temporal context information and initial output of video annotation.
- Combining multiple clusterings using fast simulated annealing4
This paper presents a fast simulated annealing framework for combining multiple clusterings based on agreement measures between partitions, which are originally used to evaluate a clustering algorithm and can achieve promising results.
- Combining latent semantic learning and reduced hypergraph learning for semi-supervised image categorization4
A semantics-aware representation is proposed which can learn latent semantics automatically from a large vocabulary of abundant visual keywords through contextual spectral embedding and a reduced hypergraph-based semi-supervised learning method is developed to exploit both labeled and unlabeled images for image categorization.
- Using Earth Mover’s Distance for Audio Clip Retrieval4
Experimental results show that the proposed approach for audio clip retrieval is better than some existing methods in terms of retrieval and ranking capabilities.
- Interact-Custom: Customized Human Object Interaction Image Generation3
A two-stage model Interact-Custom is designed, which firstly explicitly model the spatial configuration by generating a foreground mask depicting the interaction behavior, then under the guidance of this mask, the target human object interacting while preserving their identities features.
- 3
- Investigating Domain Gaps for Indoor 3D Object Detection3
This paper considers the task of adapting indoor 3D object detectors from one dataset to another, presenting a comprehensive benchmark with ScanNet, SUN RGB-D and 3D Front datasets, as well as the authors' newly proposed large-scale datasets ProcTHOR-OD and ProcFront generated by a 3D simulator.
- 3
- 3
- Uncover the Body: Occluded Person Re-identification via Masked Image Modeling3
This paper proposes an Uncover the Body Network (UBN) which exhibits the ability to remove the occlusion and attempt to restore the full body of a pedestrian and proposes a MIM (Masked Image Modeling) based method for its powerful representation of partial images to the whole.
- 3
- IEEE Access Special Section Editorial: Recent Advantages of Computer Vision3
With the emergence of big vision data and the development of artificial intelligence, it is important to investigate new theories, methods, and applications in computer vision.
- Logo Recognition via Improved Topological Constraint3
It is proven in this paper that the proposed constraint can remove the keypoints which are not on the same planar surface with the others from the k nearest neighbors of a keypoint, which is very important to logo recognition because logos are planar objects in real world.
- 3
- Visual Vocabulary Optimization with Spatial Context for Image Annotation and Classification3
A novel method is proposed that employs spatial information as side information to constrain the construction of visual vocabulary, instead of simply assigning keypoints to the nearest cluster centers, which also takes the spatial context of keypoints into consideration in the clustering process.
- Refining video annotation by exploiting inter-shot context3
This paper proposes a new approach to refine video annotation by exploiting the inter-shot context, which model the sequence of shots in video as a conditional random field with chain structure to improve the annotation accuracy.
- Story-Based Retrieval by Learning and Measuring the Concept-Based and Content-Based Similarity3
The model construction of learning classifier for story-based retrieval is constructed as follows, and the results show the approach can achieve superior performance on TRECVID 2005 benchmark.
- A Measure Based on Optimal Matching in Graph Theory for Document Similarity3
This paper proposes a new similarity measure based on optimal matching in graph theory that takes into account the structural information of a document by considering the word distributions over different text segments.
- Large-Scale Pre-Trained Models Empowering Phrase Generalization in Temporal Sentence Localization2
A phrase-level Temporal Relationship Mining (TRM) framework that employs the temporal relationship between the phrase and the whole sentence to better understand each semantic entity in the sentence to improve phrase-level prediction quality and use phrase-level predictions to refine sentence-level ones.
- Multi-Resolution Alignment for Voxel Sparsity in Camera-Based 3D Semantic Scene Completion2
A Multi-Resolution Alignment (MRA) approach to mitigate voxel sparsity in camera-based 3D semantic scene completion, which exploits the scene and instance level alignment across multi-resolution 3D features as auxiliary supervision.
- 2
- SPHERE: Semantic-PHysical Engaged REpresentation for 3D Semantic Scene Completion2
The Semantic-PHysical Engaged REpresentation (SPHERE) for camera-based SSC is proposed, which integrates voxel and Gaussian representations for joint exploitation of semantic and physical information and generates SSC results with realistic details.
- 2
- Team PKU-WICT-MIPL PIC Makeup Temporal Video Grounding Challenge 2022 Technical Report2
A phrase relationship mining framework to exploit the temporal localization relationship relevant to the fine-grained phrase and the whole sentence and to constrain the localization results of different step sentence queries to not overlap with each other through a dynamic programming algorithm is proposed.
- RCE-HIL2
HIL recognizes entailment relationships via cross-media joint inference, from image-text premises to text hypotheses, through heterogeneous interactive learning (HIL) approach, and proves the effectiveness of HIL and the intrinsic inter-media complementarity in reasoning.
- Multi-attention Guided Activation Propagation in CNNs2
A multi-attention guided activation propagation approach (MAAP), which can be applied into existing CNNs to promote their performance, and demonstrates that the applications of MAAP achieve better performance than state-of-the-art CNNs.
- Recognition of Text in Video Based on Color Clustering and Multiple Frame Integration2
This paper proposes a new approach for the text recognition of video, whose novelty mainly lies in the color-based clustering and multiple frame integration of three phases: first, in the text detection phase, the two significant features of text block are jointly considered in a video: homogeneous color, dense edges, and color- based clustering are employed to decompose the color edge map of video frame into several edge maps, which make the text Detection more accurate.
- Mining concept relationship in temporal context for effective video annotation2
A temporal model is utilized to describe the contributions of relevant concepts to the presence of a target concept in temporal context and is connected with conditional random field to obtain the refined probability of a concept occurring in the shot.
- 2
- 1
- 1
- 1
- [Reform of biology experiment teaching in the context of "Strengthening Basic Disciplines Program"-taking the course of comprehensive biology design as an example].1
The School of Life Sciences of Tianjin University has explored and practiced the reform of biology experimental teaching based on the objective of fostering biology talents in the context of "Strengthening Basic Disciplines Plan" and the STEAM education concept.
- 1
- Progressive Cross-Media Correlation Learning1
Progressive Cross-media Correlation Learning is proposed, which takes a large-scale cross-media dataset with general knowledge to guide the correlation learning on another small-scale dataset (target data) via the progressive sample selection mechanism, and outperforms 9 state-of-the-art methods.
- Coarse Label Refined Knowledge Reasoning for Fine-Grained Visual Categorization1
This work proposes a Coarse Label Refined Knowledge Reasoning (CLRKR) approach for fine-grained visual categorization, which achieves the best performance according to the experimental results on the widely-used CUB-200-2011 dataset.
- Cost-Sensitive Deep Metric Learning for Fine-Grained Image Classification1
A cost-sensitive deep metric learning (CDML) approach is proposed, which integrates confusion drivenDeep metric learning and weighted softmax for learning the difference among the confusing subcategories.
- Attention-Sharing Correlation Learning for Cross-Media Retrieval1
Inspired by attention mechanism, this paper proposes the attention-sharing correlation learning network, which is an end-to-end network to generate cross-media common representation for retrieval, which can not only emphasize the single-media discriminative parts, but also enhance the cross- media fine-grained consistent pattern, and so learn more precious cross- Media correlation to improve retrieval accuracy.
- Cross-media residual correlation learning1
Rival correlation network (RCN) optimizes common representation learning with a residual function, which can fit the optimal mapping from separate representation to common representation and relieve the multi-objective optimization problem.
- Vocabulary hierarchy optimisation based on spatial context and category information1
An object function is proposed to describe the consistence of visual words, with category information of images and spatial context of keypoints, and then simulated annealing algorithm is adopted to search for a sub-optimal solution, which corresponds to a visual vocabulary selected from the vocabulary tree.
- PDSS1
The experiment adopts the extensively-used Labeled Face in the Wild (LFW) unconstrained face recognition dataset (13K faces), and the proposed PDSS approach achieves the best result, compared with the state-of-the-art methods.
- PKU-ICST at TRECVID 2012: Known-item Search Task1
Three kinds of text information are adopted, which are XML documents, ASR and OCR, and they are index and search the three kinds of pre-processed text individually with Lucene and combined and re-ranked by two re-ranking approaches.
- Color-Based Text Extraction for the Image1
This paper adaptively selects the relatively best color plane for the binarization of text region image, which uses the text contrast difference among the color planes, and considers the color difference between the text strokes and noises.
- A New Re-ranking Method for Generic Chinese Text Summarization and Its Evaluation1
A new EMD-MMR (EMD: earth mover's distance; MMR: maximal marginal relevance) re-ranking method is proposed for generic Chinese text summarization and the performance of the proposed approach with a novel nk-blind method is evaluated.
- –
- –
- [Exploration and practice of cultivating top-notch innovative talents in biological sciences under the dual-track framework integrating "Strengthening Basic Disciplines Program" and "Top-notch Student Training Program"].–
The School of Life Sciences at Tianjin University has established replicable reform pathways for cultivating top-notch innovative talents in biological sciences at domestic universities, while also contributing fresh perspectives to the high-quality development of life science education in the new era.
- –
- –
- –
- –
- –
- –
- Bi-C 2 R: Bidirectional Continual Compatible Representation for Re-Indexing Free Lifelong Person Re-Identification–
A Bidirectional Continuous Compatible Representation (Bi-C) framework to continuously update the gallery features extracted by the old model to perform efficient L-ReID in a compatible manner and a feature-level exponential moving average strategy is designed to adaptively fill the diverse knowledge gaps between different data domains.
- –
- Enhancing video quality assessment through brightness information–
This paper proposes a novel VQA method grounded in the HVS that introduces brightness information from video frames as embeddings, thereby enhancing the model's alignment with the HVS and improving efficiency.
- –
- –
- –
- –
- Enhancing Unsupervised Audio Representation Learning via Adversarial Sample Generation–
This paper proposes a new self-supervised learning mechanism, which enhances the audio representation by first generating adversarial samples, then driving CNN to distinguish the embeddings of negative pairs in the latent space.
- –
- Hologram QSAR and Topomer CoMFA Study on Naphthyridone Derivatives as ATAD2 Bromodomain Inhibitors–
The QSAR analyses of naphthyridone derivatives conducted using hologram quantitative structure–activity relationship (HQSAR) and topomer comparative molecular field analysis (topomer CoMFA) provide useful clues for the design of potential ATAD2 bromodomain inhibitors.
- –
- –
- –
- –
- –
- –
- –
- –
- Web video search by mutual boosting between the inside and outside text of video–
This paper proposed a new approach for automatic caption detection and extraction, which boosts mutually and utilizes jointly the inside and outside text of video to support the video-based and frame-based search for web video.
- Background Modeling under Free Moving Camera Environment–
A novel Multi-layer Homography algorithm for background modeling under free moving camera environment is proposed and experiment shows it is effective for background modeled under freeMoving camera environments.
- A Theoretical Approach to Construct Highly Discriminative Features with Application in AdaBoost–
A theoretical approach to construct highly discriminative features, which is named composed features, from Haar-like features is proposed, employed to train a multi-view face detector.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-15. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.