A5032611421- Human Pose and Action Recognition
- Video Surveillance and Tracking Methods
- Gait Recognition and Analysis
- Face recognition and analysis
- Face and Expression Recognition
- Anomaly Detection Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- Multimodal Machine Learning Applications
- Medical Image Segmentation Techniques
- Remote-Sensing Image Classification
- Infrastructure Maintenance and Monitoring
- Image Retrieval and Classification Techniques
- Sparse and Compressive Sensing Techniques
- Robotics and Sensor-Based Localization
- Asphalt Pavement Performance Evaluation
- Remote Sensing and Land Use
- Concrete Corrosion and Durability
- Digital Media Forensic Detection
- Gene expression and cancer classification
- Advanced Vision and Imaging
- Advanced Neural Network Applications
- Cell Image Analysis Techniques
- Advanced Clustering Algorithms Research
- Biometric Identification and Security
- 3D Shape Modeling and Analysis
Chongqing University
2016-2025
Dalian Institute of Chemical Physics
2023-2024
Chinese Academy of Sciences
2023-2024
University of Chinese Academy of Sciences
2023-2024
Dalian National Laboratory for Clean Energy
2023
Ministry of Education of the People's Republic of China
2013-2015
Northeast Normal University
2008
National University of Singapore
2005-2007
Shanghai First People's Hospital
2000-2002
The inert C(sp3)–H bond and easy overoxidation of toluene make the selective oxidation to benzaldehyde a great challenge. Herein, we present that photocatalyst, constructed with small amount (1 mol %) amorphous BiOCl nanosheets assembled on TiO2 (denoted as 0.01BOC/TiO2), shows excellent performance in benzaldehyde, 85% selectivity at 10% conversion, formation rate is up 1.7 mmol g–1 h–1, which 5.6 3.7 times bare BOC, respectively. In addition charge separation function BOC/TiO2...
Semi-supervised learning based on consistency offers significant promise for enhancing medical image segmentation. Current approaches use copy-paste as an effective data perturbation technique to facilitate weak-to-strong learning. However, these techniques often lead a decrease in the accuracy of synthetic labels corresponding and introduce excessive perturbations distribution training data. Such over-perturbation causes stray from its true distribution, thereby impairing model's...
Recently, clustering algorithms based on deep AutoEncoder attract lots of attention due to their excellent performance. On the other hand, success PCA-Kmeans and spectral corroborates that orthogonality embedding is beneficial increase accuracy. In this paper, we propose a novel dimensional reduction model, called Orthogonal (OAE), which encourages learned embedding. Furthermore, joint Clustering framework (COAE), new capable extracting latent predicting assignment simultaneously. The COAE...
Deep learning-based pavement cracks detection methods often require large-scale labels with detailed crack location information to learn accurate predictions. In practice, however, locations are very difficult be manually annotated due various visual patterns of crack. this paper, we propose a Domain Adaptation-based Crack Detection Network (DDACDN), which learns domain invariant features by taking advantage the source knowledge predict multi-category in target domain, where only image-level...
Applying deep learning to predict patient prognostic survival outcomes using histological whole-slide images (WSIs) and genomic data is challenging due the morphological transcriptomic heterogeneity present in tumor microenvironment. Existing learning-enabled methods often exhibit biases, primarily because knowledge used guide directional feature extraction from WSIs may be irrelevant or incomplete. This results a suboptimal sometimes myopic understanding of overall pathological landscape,...
We investigate in this paper the problem of face verification presence makeups. To our knowledge, has less formally addressed literature. A key challenge is how to increase measured similarity between images same person without and with In paper, we propose a novel approach for makeup-robust verification, by measuring correlations meta subspace. The subspace learned using canonical correlation analysis (CCA), objective that intra-personal sample are maximized. Subsequently, discriminative...
This paper focuses on RGB-infrared person re-identification, which is challenged by a large modality gap between RGB and infrared images. Most existing methods attempt to learn discriminative modality-invariant features. These make use of identity annotations while they do not sufficiently exploit intra-modality cross-modality sample relations using annotations. In this paper, we propose Cross-modality channel Mixup Modality Decorrelation method (CMMD) that explores at both image feature...
With video-level labels, weakly supervised temporal action localization (WTAL) applies a localization-by-classification paradigm to detect and classify the in untrimmed videos. Due characteristic of classification, class-specific background snippets are inevitably mis-activated improve discriminability classifier WTAL. To alleviate disturbance background, existing methods try enlarge discrepancy between through modeling with pseudo-snippet-level annotations, which largely rely on artificial...
Person re-identification (re-ID) is a challenging problem in the community which aims at identifying person surveillance video. Despite recent advance field of computer vision, re-ID still presents great challenge since person's presence various under different illumination, viewpoints, occlusion, and background clutter. In this paper, to exploit more discriminative information appearance, we propose novel pose invariant deep metric learning (PIDML) method an improved triplet loss for re-ID....