A5100455049- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Advanced Optical Sensing Technologies
- Photoacoustic and Ultrasonic Imaging
- Parallel Computing and Optimization Techniques
- Domain Adaptation and Few-Shot Learning
- Industrial Vision Systems and Defect Detection
- Image Enhancement Techniques
- Ocular and Laser Science Research
- Energy Load and Power Forecasting
- Stochastic Gradient Optimization Techniques
- 3D Surveying and Cultural Heritage
- Distributed and Parallel Computing Systems
- Neural Networks and Applications
- Brain Tumor Detection and Classification
- Machine Learning and ELM
- Video Surveillance and Tracking Methods
- Electric Power System Optimization
- Robotics and Sensor-Based Localization
- Power Systems and Renewable Energy
- Video Analysis and Summarization
- Spacecraft and Cryogenic Technologies
- Machine Learning and Data Classification
- Random lasers and scattering media
- Remote-Sensing Image Classification
Harbin Institute of Technology
2023-2025
Weihai Science and Technology Bureau
2025
Ministry of Industry and Information Technology
2023-2024
Shandong Marine Resource and Environment Research Institute
2023-2024
Beijing University of Posts and Telecommunications
2020-2021
National University of Defense Technology
2015-2020
National Supercomputing Center in Shenzhen
2018
Institute of Software
2017
Changsha University
2015
For the demand for long-range and high-resolution target reconstruction of slow-moving small underwater targets, research on single-photon lidar technology is being carried out. This paper reports sequential multimodal adaptive algorithm based spatiotemporal sequence fusion, which has strong information extraction noise filtering ability can reconstruct depth reflective intensity from complex echo photon time counts spatial pixel relationships. The method consists three steps: data...
Ship detection based on remote sensing images holds significant importance in both military and economic domains. Ships within such exhibit diverse scales, dense distributions, arbitrary orientations, narrow shapes, which pose challenges for accurate recognition. This paper introduces an improved S2A-Net (Single-shot Alignment Network) oriented object algorithm ship detection. In network structure, pyramid squeeze attention is embedded order to focus key features a context information module...
Deep learning (DL) is currently the most promising approach in complicated applications such as computer vision and natural language processing. It thrives with large neural networks datasets. However, larger models datasets result longer training times that impede research development progress. The modern high-performance data-parallel nature of hardware equipped high computing power, GPUs, has triggered widespread adoption DL frameworks, Caffe, Torch, TensorFlow. frameworks cannot make...
Traditional LiDAR and air-medium-based single-photon struggle to perform effectively in high-scattering environments. The laser beams are subject severe medium absorption multiple scattering phenomena such conditions, greatly limiting the maximum operational range imaging quality of system. high sensitivity temporal resolution enable high-resolution depth information acquisition under limited illumination power, making it highly suitable for operation environments with extremely poor...
Training deep learning (DL) is a computationally intensive process; as result, training time can become so long that it impedes the development of DL. High performance computing clusters, especially supercomputers, are equipped with large amount resources, storage and efficient interconnection ability, which train DL networks better faster. In this paper, we propose method to distributed high efficiency. First, hierarchical synchronous Stochastic Gradient Descent (SGD) strategy, make full...
Heterogeneous computing has been used widely since accelerators like Graphic Processing Unit (GPU) and Intel Many Integrated Core (MIC) can offer an order of magnitude higher compute power for arithmetic intensive data-parallel workloads. However, heterogeneous programming is more complicated there no shared memory between CPU MIC. Programmers must distinguish the local or remote access data transmit MIC explicitly. Furthermore, standard offload models Language Extensions Offload (LEO)...
The random forest algorithm is an ensemble classifier based on the decision tree model. It has a wide range of applications in machine learning, data mining and other fields. With emergence big age, training process becomes very lengthy. Most studies speed up forests through small clusters or high performance device, however, few pay attention to supercomputers. In this paper, we propose parallel framework supercomputers called DistForest which can utilize multiple nodes train with large...
Abstract Fully Convolution Network and its following works has achieved the state-of-art performance on task of RGB semantic segmentation. However, there still lacks an effective method to sufficiently leverage geometric information depth image accomplish RGB-D To this end, paper proposed a new containing two parts: 1) simple but useful way based Fast Marching Method inpaint area no-measured-depth pixels, which produces better result than standard dataset we used in paper; 2) fusion...
In this paper, we build a lightweight convolution neural network M-SOSAnet that combines efficiency and accuracy for edge devices. Just as DenseNet connects each layer to every other in the network. Although dense connection can effectively keep information between middle layers, increasing input channels by leads resource consumption, which greatly increases amount of parameters is inefficient. MobileNet series use depthwise separable convolutions, reduces calculation parameters, but will...