A5047341940- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Marine Bivalve and Aquaculture Studies
- Image Retrieval and Classification Techniques
- Human Pose and Action Recognition
- Face recognition and analysis
- 3D Shape Modeling and Analysis
- Aquaculture disease management and microbiota
- Image Enhancement Techniques
- Face and Expression Recognition
- Video Analysis and Summarization
- Advanced Image Processing Techniques
- 3D Surveying and Cultural Heritage
- Advanced Vision and Imaging
- Sparse and Compressive Sensing Techniques
- Aquaculture Nutrition and Growth
- Advanced Data Storage Technologies
- Direction-of-Arrival Estimation Techniques
- Gait Recognition and Analysis
- Genetic diversity and population structure
- Distributed and Parallel Computing Systems
- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Handwritten Text Recognition Techniques
Southwest University
2016-2025
Third Affiliated Hospital of Guangzhou Medical University
2025
Guangzhou Medical University
2025
Zhejiang Mariculture Research Institute
2015-2024
Zhejiang Ocean University
2022
Citrus Research Institute
2022
Shanghai Ocean University
2022
Chinese Academy of Fishery Sciences
2017-2018
James Cook University
2015
Huanghuai University
2010-2013
Feature representation learning is a key component in 3D point cloud analysis. However, the powerful convolutional neural networks (CNNs) cannot be applied due to irregular structure of clouds. Therefore, following tremendous success transformer natural language processing and image understanding tasks, this paper, we present novel architecture, named Dual Transformer Network (DTNet), which mainly consists Point Cloud (DPCT) module. Specifically, by aggregating well-designed point-wise...
The previous deep CNN-based single-image dehazing methods are devoted to improving the performance by increasing network's depth and width. In this paper, a novel Large Kernel Convolution Dehaze Network (LKD-Net) is proposed enhance size of convolutional kernel. main module in LKD-Net designed Block (LKD Block), which consists Decomposition deep-wise (DLKCB) Channel Enhanced Feed-forward (CEFN). DLKCB reduce massive amount computational overhead parameters large kernel splitting convolution...
Effective and efficient 3D semantic segmentation from large-scale LiDAR point cloud is a fundamental problem in the field of autonomous driving. In this paper, we present Transformer-Range-View Network (TransRVNet), novel powerful projection-based CNN-Transformer architecture to infer point-wise semantics. First, Multi Residual Channel Interaction Attention Module (MRCIAM) introduced capture channel-level multi-scale feature model intra-channel, inter-channel correlations based on attention...
Accurate and fast scene understanding is one of the chal-lenging task for autonomous driving, which requires to take full advantage LiDAR point clouds semantic segmen-tation. In this paper, we present a concise efficient image-based segmentation network, named CENet. order improve descriptive power learned features reduce computational as well time complex-ity, our CEN et integrates convolution with larger ker-nel size instead MLP, carefully-selected activation functions, multiple auxiliary...
Coastal aquaculture plays an important role in the provision of seafood, sustainable development regional and global economy, protection coastal ecosystems. Inappropriate planning disordered intensive may cause serious environmental problems socioeconomic losses. Precise delineation classification different kinds areas are vital for management. It is difficult to extract using conventional spectrum, shape, or texture information. Here, we proposed object-based method combining multi-scale...
Self-attention mechanism recently achieves impressive advancement in Natural Language Processing (NLP) and Image domains. Its permutation invariance property makes it ideally suitable for point cloud processing. Inspired by this remarkable success, we propose an end-to-end architecture, dubbed Cross-Level Cross-Scale Cross-Attention Network (3CROSSNet), representation learning. First, a point-wise feature pyramid module is introduced to hierarchically extract features from different scales...
The blood clam, Tegillarca granosa, is widely distributed along the coasts of Indo-Pacific region, providing an excellent opportunity to study gene flow in sessile marine mollusks. In present study, amplified fragment length polymorphism (AFLP) DNA markers were used analyze genetic structure five clam populations. Genetic differentiation (Gst) and Nei's distances between population pairs found range from 0.0245 0.0785 0.0398 0.1125, respectively. An AMOVA analysis showed that 89.09%...
The mechanistic role of insulin resistance(IR) between obesity and Depressive Symptoms has long been debated. This study aims to explore the relationship BMI investigate mediating IR in this association. We rigorously selected data from National Health Nutrition Examination Survey (NHANES) 2005 2018. Multivariate logistic regression analysis, Restricted cubic splines (RCS) Subgroup analysis were employed assess correlation Symptoms. Mediation was conducted examine Overweight/obesity After...
The paper presents an adaptive wear-leveling scheme based on several wear-thresholds in different periods. basic idea behind this is that blocks can have wear-out speeds and the mechanism does not conduct data migration until erasure counts of some hot hit a threshold. Through series emulation experiments realistic disk traces, we show proposed reduce total yield uniform among all at late lifetime storage devices. As result, only performance systems be advanced, lifespan flash-based memory...
This paper presents an initiative data prefetching scheme on the storage servers in distributed file systems for cloud computing. In this technique, client machines are not substantially involved process of prefetching, but can directly prefetch after analyzing history disk I/O access events, and then send prefetched to relevant proactively. To put technique work, information about nodes is piggybacked onto real requests, forwarded server. Next, two prediction algorithms have been proposed...
Background Blood clams (Tegillarca granosa) are one of the most commercial shellfish in China and South Asia with wide distribution Indo-Pacific tropical to temperate estuaries. However, recent data indicate a decline germplasm this species. Furthermore, molecular mechanisms underpinning reproductive regulation remain unclear information regarding genetic diversity is limited. Understanding biology important interpreting their embryology development, reproduction population structure....