A5101919507- Magnetic confinement fusion research
- Advanced Neural Network Applications
- Fusion materials and technologies
- Superconducting Materials and Applications
- Advanced Image and Video Retrieval Techniques
- Laser-Plasma Interactions and Diagnostics
- Visual Attention and Saliency Detection
- Ionosphere and magnetosphere dynamics
- Particle accelerators and beam dynamics
- Human Mobility and Location-Based Analysis
- Land Use and Ecosystem Services
- Advanced Vision and Imaging
- Generative Adversarial Networks and Image Synthesis
- Advanced Measurement and Detection Methods
- 3D Modeling in Geospatial Applications
- Multimodal Machine Learning Applications
- Image and Object Detection Techniques
- Infrared Target Detection Methodologies
- Sentiment Analysis and Opinion Mining
- Complex Network Analysis Techniques
- Data Management and Algorithms
- Video Surveillance and Tracking Methods
- Thermography and Photoacoustic Techniques
- Geographic Information Systems Studies
- Advanced Image Processing Techniques
Tsinghua–Berkeley Shenzhen Institute
2022-2024
Dalian University of Technology
2024
Zhejiang University of Finance and Economics
2020-2023
Tongling University
2018-2021
Tsinghua University
2021
Donghua University
2014-2016
Institute of Plasma Physics
2016
Chinese Academy of Sciences
2016
Knowledge distillation has been applied to image classification successfully. However, object detection is much more sophisticated and most knowledge methods have failed on it. In this paper, we point out that in detection, the features of teacher student vary greatly different areas, especially foreground background. If distill them equally, uneven differences between feature maps will negatively affect distillation. Thus, propose Focal Global Distillation (FGD). separates background,...
This paper provides an investigation of the production/transport properties W impurity in edge experimental advanced superconducting tokamak (EAST) H-mode discharges with upper-single-null configurations by using DIVIMP Monte-Carlo code. The background plasmas are provided SOLPS5.0 calculations. Firstly, to address detailed dependence behaviors on plasma conditions scraped-off layer (SOL)/divertor region EAST, two pre-edge localized mode (ELM) cases divertor high-recycling and...
To support the ITER early non-nuclear operational phase, recent Helium experiments have been investigated intensively in superconducting tokamak EAST equipped with ITER-like Tungsten divertor, under condition of low injected torque. This paper demonstrates experimental observations from these experiments. In particular, we report impact concentration on L-H transition power threshold, energy confinement, and ELM behavior as well. Dependences threshold electron density, plasma current...
Scatterplots are always employed to visualize geographical point datasets, which often suffer from an overdraw problem due the increase of data sizes. A variety sampling strategies have been proposed reduce and visual clutter with spatial densities points taken into account. However, informative attributes associated also play significant roles in exploration datasets. In this paper, we propose attribute-based abstraction method simplify cluttered visualization large-scale points. Spatial...
The controlling of plasma shape and position are essential to the success Tokamak discharge. A real-time image acquisition system was designed obtain radiation during discharge processes in Experimental Advanced Superconducting (EAST) device. hardware structure software design this visible camera introduced detail. According general EAST layout observation window, spatial location discharging measured. An improved Sobel edge detection algorithm using iterative threshold proposed detect...
Detailed investigations on the filamentary structures associated with type-I edge-localized modes (ELMs) should be helpful for protecting materials of a plasma-facing wall future large device. Related experiments have been carefully conducted in Experimental Advanced Superconducting Tokamak (EAST) using combined Langmuir-magnetic probes. The experimental results indicate that radially outward velocity ELMy filaments can up to 1.7 km s−1 far scrape-off layer (SOL) region. It is remarkable...
A tangential fast visible camera has been set up in EAST tokamak for the study of edge MHD instabilities such as ELM. To determine 3-D information from CCD images, Tsai's two-stage technique was utilized to calibrate high-speed imaging system ELM study. By applying tiles passive stabilizers device calibration pattern, transformation parameters transforming a world coordinate 2-D image were obtained, including rotation matrix, translation vector, focal length and lens distortion. The errors...
Abstract Addressing the challenge of ensuring robustness in vision‐based target recognition algorithms under adverse weather conditions, such as rain, snow, and fog, is crucial. In this paper, we introduce a novel approach for road detection that can effectively operate various conditions. Our method based on cascade task framework detection, complemented by image restoration techniques. Specifically, have developed denoising network tailored to meet demands de‐raining snow removal tasks....
The heat fluxes on the lower outer divertor plate were calculated using DFLUX code with surface temperature measured by IR camera experimental advanced superconducting tokamak. analyzed discharges single-null configuration different auxiliary heating: hybrid wave current drive (LHCD) combined neutral beam injection (NBI) and LHCD ion cyclotron resonant heating (ICRH). of target was about 170 °C in + NBI H-mode, high mostly distributed location near joint dome plate. increased as turned...
Knowledge distillation has been applied to image classification successfully. However, object detection is much more sophisticated and most knowledge methods have failed on it. In this paper, we point out that in detection, the features of teacher student vary greatly different areas, especially foreground background. If distill them equally, uneven differences between feature maps will negatively affect distillation. Thus, propose Focal Global Distillation (FGD). separates background,...
Abstract With the increasing size of social media datasets, it has become a difficult task to explore significant features such as sentiments from traditional geographical visualizations, because visual elements data items always overlap heavily with each other. A large number sampling methods have been proposed reduce clutter and simplify visualizations. However, sentiment are ignored in course which make easily misunderstand across local areas. In this paper, we propose abstraction...