A5018564842- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Human Pose and Action Recognition
- Advanced Image and Video Retrieval Techniques
- Oral microbiology and periodontitis research
- COVID-19 diagnosis using AI
- Bone Metabolism and Diseases
- Gait Recognition and Analysis
- Face recognition and analysis
- Machine Learning and ELM
- Salivary Gland Disorders and Functions
- Video Analysis and Summarization
- Cancer-related molecular mechanisms research
- Anomaly Detection Techniques and Applications
- Medicinal Plant Pharmacodynamics Research
- Advanced Vision and Imaging
- Photodynamic Therapy Research Studies
- Bone health and treatments
- Dental Health and Care Utilization
- Speech and Audio Processing
- Oral Health Pathology and Treatment
- Music and Audio Processing
- Topic Modeling
Wuhan University
2021-2025
Institute of Automation
2018-2024
Chinese Academy of Sciences
2016-2024
Shandong Institute of Automation
2019-2023
Ministry of Science and Technology of the People's Republic of China
2022
Stomatology Hospital
2022
University of Chinese Academy of Sciences
2016-2020
University of Science and Technology of China
2020
National Institute of Measurement and Testing Technology
2019
Institute of Computing Technology
2016-2017
The huge variance of human pose and the misalignment detected images significantly increase difficulty person Re-Identification (Re-ID). Moreover, efficient Re-ID systems are required to cope with massive visual data being produced by video surveillance systems. Targeting solve these problems, this work proposes a Global-Local-Alignment Descriptor (GLAD) an indexing retrieval framework, respectively. GLAD explicitly leverages local global cues in body generate discriminative robust...
Learning discriminative representations for unseen person images is critical Re-Identification (ReID). Most of current approaches learn deep in classification tasks, which essentially minimize the empirical risk on training set. As shown our experiments, such commonly focus several body parts to set, rather than entire human body. Inspired by structural minimization principle SVM, we revise traditional representation learning procedure both and risk. The evaluated proposed part loss,...
Generalized zero-shot learning aims to recognize images from seen and unseen domains. Recent methods focus on a unified semantic-aligned visual representation transfer knowledge between two domains, while ignoring the effect of semantic-free in alleviating biased recognition problem. In this paper, we propose novel Domain-aware Visual Bias Eliminating (DVBE) network that constructs complementary representations, i.e., semantic-aligned, treat domains separately. Specifically, explore...
Lung cancer is the leading cause of mortality, and early detection key to improving survival. However, there are no reliable blood-based tests currently available for early-stage lung diagnosis. Here, we performed single-cell RNA sequencing different cancers found that lipid metabolism was broadly dysregulated in cell types, with glycerophospholipid as most altered metabolism–related pathway. Untargeted lipidomics carried out an exploratory cohort 311 participants. Through support vector...
Prompt tuning is an effective way to adapt the pretrained visual-language model (VLM) downstream task using task-related textual tokens. Representative CoOp-based work combines learnable tokens with class obtain specific knowledge. However, knowledge worse generalization unseen classes because it forgets essential general having a strong ability. To tackle this issue, we introduce novel Knowledge-guided Context Optimization (KgCoOp) enhance ability of prompt for classes. The key insight...
The huge variance of human pose and the misalign-ment detected images significantly increase difficulty pedestrian image matching in person Re-Identification (Re-ID). Moreover, massive visual data being produced by surveillance video cameras requires highly efficient Re-ID systems. Targeting to solve first problem, this work proposes a robust discriminative descriptor, namely, Global-Local-Alignment Descriptor (GLAD). For second treats as retrieval an indexing framework. GLAD explicitly...
Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal cancers, characterized by rapid progression, metastasis, and difficulty in diagnosis. However, there are no effective liquid-based testing methods available for PDAC detection. Here we introduce a minimally invasive approach that uses machine learning (ML) lipidomics to detect PDAC. Through greedy algorithm mass spectrum feature selection, optimized 17 characteristic metabolites as detection features developed liquid...
Multi-source domain adaptation (MSDA) aims to transfer knowledge from multi-source domains one target domain. Inspired by single-source adaptation, existing methods solve MSDA aligning the data distributions between and each source However, with dissimilar would harm representation learning. To address above issue, an intuitive motivation of is using attention mechanism enhance positive effects similar domains, suppress negative domains. Therefore, we propose Attention-Based Multi-Source...
Early childhood caries (ECC) is a public healthcare concern that greatly reduces the quality of life young children. As leading factor ECC, cariogenic biofilms are composed acidogenic/aciduric pathogens and extracellular polysaccharides (EPSs), creating an acidic protected microenvironment. Antimicrobial photodynamic therapy (aPDT) noninvasive, painless, efficient therapeutic approach suitable for treating ECC. However, due to hyperfine structure biofilms, most photosensitizers (PSs) could...
Recently, Multiple Object Tracking has achieved great success, which consists of object detection, feature embedding, and identity association. Existing methods apply the three-step or two-step paradigm to generate robust trajectories, where association is independent other components. However, results in identity-aware knowledge contained tracklet not be used boost detection embedding modules. To overcome limitations existing methods, we introduce a novel Unified Model (UTM) bridge those...
Recent years have witnessed the significant advance in fine-grained visual categorization, which targets to classify objects belonging same species. To capture enough subtle differences and build discriminative description, most of existing methods heavily rely on artificial part annotations, are expensive collect real applications. Motivated conquer this issue, paper proposes a multi-level coarse-to-fine object description. This novel description only requires original image as input, but...
To analyse the characteristics of oral microbiomes and expected to find biomarkers about Alzheimer's disease (AD).AD patients (n = 26) cognitive intact people were examined for cognition, depression, health collected saliva gingival crevicular fluid (GCF) in morning. Full-length 16S rRNA gene was amplified sequencing performed using PacBio platform.The predominant bacterium salivary microbiome periodontal from AD Streptococcus oralis Porphyromonas gingivalis, respectively. With respect β...
Periodontitis is a worldwide oral disease induced by the interaction of subgingival bacteria and host response characterized local inflammation, bone resorption, tooth loss. Ginsenoside Rd (Rd) biologically active component derived from Panax ginseng has been demonstrated to exert antibacterial anti-inflammatory activities. This study aims investigate inhibitory efficiency towards Porphyromonas gingivalis ( P. ), periodontal inflammatory response, osteoclastogenesis in vitro further validate...
Prompt tuning is a valuable technique for adapting visual language models (VLMs) to different downstream tasks, such as domain generalization and learning from few examples. Previous methods have utilized Context Optimization approaches deduce domain-shared or cross-modality prompt tokens, which enhance discriminative ability in textual contexts. However, these inferred training data, cannot adapt perfectly the distribution of test dataset. This work introduces novel approach called...
Compared with traditional image classification, fine-grained visual categorization is a more challenging task, because it targets to classify objects belonging the same species, e.g., hundreds of birds or cars. In past several years, researchers have made many achievements on this topic. However, most them are heavily dependent artificial annotations, bounding boxes, part and so on. The requirement annotations largely hinders scalability application. Motivated release such dependence, paper...
Person search targets to the probe person from unconstrainted scene images, which can be treated as combination of detection and matching. However, existing methods based on Detection-Matching framework ignore objectness repulsion (OR) are both beneficial reduce effect distractor images. In this paper, we propose an OR similarity by jointly considering information. Besides traditional visual term, also contains term a term. The images that not contain boost performance improving ranking...
In the past years, pedestrian detection has achieved significant progress via improving visual description. However, description is not robust to discover occluded pedestrian, which bottleneck of existing methods. Targeting overcome shortcoming description, we employ human pose information, complementary address occlusion and false positive failure problems in detection. The advantage using information that estimation model can localize local part once occluded. By embedding with propose a...
Multi-target multi-camera tracking (MTMCT) targets to generate trajectories of the object that appeared under multiple cameras automatically. MTMCT can be treated as a combination intra-camera and cross-camera tracking. The existing work only employs global description perform tracklet generating. However, cannot model local similarity between targets, leading methods not robust occlusion fast motion. To handle mentioned problem, we propose an online Optical-based Pose Association (OPA) for...
Domain adaptation aims to transfer the knowledge learned from a labeled source domain an unlabeled target domain, which has different data distribution with domain. Most of existing methods focus on aligning between and domains but ignore discrimination feature space among categories, leading samples close decision boundary be misclassified easily. To address above issue, we propose Margin-based Adversarial Joint Alignment (MAJA) constrain spaces aligned discriminative. The proposed MAJA...
Recently, unsupervised domain adaptation in person re-identification (ReID) has been widely studied to improve the generalization ability of ReID model. Some existing methods focus on handling intra-domain image variations caused by different camera configurations, pose, illumination, and background target domain. However, they fail fully mine underlying consistency constraints contained unlabeled dataset. To comprehensively investigate for representation learning, we introduce two deal with...
Bilinear pooling achieves great success in fine-grained visual recognition (FGVC). Recent methods have shown that the matrix power normalization can stabilize second-order information bilinear features, but some problems, e.g., redundant and over-fitting, remain to be resolved. In this paper, we propose an efficient Multi-Objective Matrix Normalization (MOMN) method simultaneously normalize a representation terms of square-root, low-rank, sparsity. These three regularizers not only...