A5047723747- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Industrial Vision Systems and Defect Detection
- Human Pose and Action Recognition
- Visual Attention and Saliency Detection
- Anomaly Detection Techniques and Applications
- Gait Recognition and Analysis
- Infrastructure Maintenance and Monitoring
- Water Systems and Optimization
- Non-Invasive Vital Sign Monitoring
- Hand Gesture Recognition Systems
- AI in cancer detection
- ECG Monitoring and Analysis
- EEG and Brain-Computer Interfaces
- Urban Stormwater Management Solutions
- Face and Expression Recognition
- Medical Image Segmentation Techniques
- Heart Rate Variability and Autonomic Control
- Emotion and Mood Recognition
- Random lasers and scattering media
- Face Recognition and Perception
- Face recognition and analysis
- Traditional Chinese Medicine Studies
- Advanced X-ray Imaging Techniques
- Non-Destructive Testing Techniques
Southeast University
2022-2025
Southeast University
2024
University of Shanghai for Science and Technology
2019-2023
Ministry of Education of the People's Republic of China
2023
Affiliated Hospital of Southwest Medical University
2018-2021
The spatial information of Electroencephalography (EEG) is essential for emotion recognition model to learn discriminative feature. convolutional networks and recurrent are the conventional choices complex dependencies through a number electrodes brain regions. However, these models have difficulty in capturing long-range due operations local feature learning. To enhance EEG improve accuracy recognition, we propose transformer- based hierarchically from electrode level brain-region-level. In...
Automatic detection of casting defects on radiography images is an important technology to automatize digital defect inspection. Traditionally, in industrial application, conventional methods are inefficient when the targets small, local, and subtle complex scenario. Meanwhile, outperformance deep learning models, such as convolutional neural network (CNN), limited by a huge volume data with precise annotations. To overcome these challenges, efficient CNN model, only trained image-level...
The temporal and spatial information of electroencephalogram (EEG) are essential for the emotion recognition model to learn discriminative features. Hence, we propose a novel hybrid spatial-temporal feature fusion neural network (STFFNN) extract features integrate complementary information. generated power topographic maps, which capture dependencies among electrodes, fed convolutional (CNN) learning. Furthermore, instance normalizations (INs) batch (BNs) within CNN appropriately combined...
An automatic vision-based sewer inspection plays a key role of sewage system in modern city. Recent advances focus on utilizing deep learning model to realize the system, benefiting from capability data-driven feature extraction. However, ambiguity defects space is ignored, deteriorating performance inspection. There are two reasons for such ambiguity. First, defect-irrelevant region interferes extraction model. Second, setting multilabel an inherent challenge extracting discriminative...
Learning to segmentation without large-scale samples is an inherent capability of human. Recently, Segment Anything Model (SAM) performs the significant zero-shot image segmentation, attracting considerable attention from computer vision community. Here, we investigate SAM for medical analysis, especially multi-phase liver tumor (MPLiTS), in terms prompts, data resolution, phases. Experimental results demonstrate that there might be a large gap between and expected performance. Fortunately,...
Unmanned aerial vehicles (UAVs) represent an essential component of advanced intelligent equipment that can be used as perception system by installing various sensors such vision, hearing, touch, taste, and smell to achieve intelligently integrated environments. However, these with environmental information may threatened internal external attacks, causing a great challenge the security UAV. The original relied on expert knowledge base prevent but weaknesses lacking proactivity flexibility...
Poly (ADP-ribose) polymerase (PARP) is a key enzyme in the repair process of DNA strand breaks (DSBs). Olaparib (Ola) PARP inhibitor that involved arresting release from radiotherapy (RT)-induced damaged to potentiate effect RT. Although underlying mechanisms for radiosensitization effects Ola are well understood vitro, vivo still unclear. Moreover, poor water solubility and severe toxicity two major impediments clinical success Ola.
Subtle variations are invisible to the naked eyes in human physiological signals can reflect important biological and health indicators. Although numerous computer vision methods have been proposed recover magnify these changes, most of them either only focus on identifying recognizing explicit features such as shapes textures, or weak long-term temporal modeling spatiotemporal interactive perception implicit biometrics. Therefore, it is difficult for robustly overcome various disturbances...
A systemic immune related response (SIME) of radiotherapy has been occasionally observed on metastatic tumors, but the clinical outcomes remain poor. Novel treatment approaches are therefore needed to improve SIME ratio. We used a combination hypo-fractionated radiation therapy (H-RT) with low-dose total body irradiation (L-TBI) in syngeneic mouse model breast and colon carcinoma. The H-RT L-TBI potentially enhanced by infiltration CD8 + T cell altering immunosuppressive microenvironment...
Light scattering is a pervasive problem in many areas. Recently, deep learning was implemented speckle reconstruction. To better investigate the key feature extraction and generalization abilities of networks for sparse pattern reconstruction, we develop "one-to-all" self-attention armed convolutional neural network (SACNN). It can extract local global properties different types patterns, unseen glass diffusers, untrained detection positions. We quantitatively analyzed performance ability...
Optical microscopes have long been essential for many scientific disciplines. However, the resolution and contrast of such microscopic images are dramatically affected by aberrations. In this study, compacted with adaptive optics, we propose a machine learning technique, called 'phase-retrieval deep convolutional neural networks (PRDCNNs)'. This aberration determination architecture is direct exhibits high accuracy certain generalisation ability. Notably, its performance surpasses those...
Recognizing the human emotion automatically from visual characteristics plays a vital role in many intelligent applications. Recently, gait-based recognition, especially gait skeletons-based characteristic, has attracted much attention, while available methods have been proposed gradually. The popular pipeline is to first extract affective features joint skeletons, and then aggregate skeleton as feature vector for classifying emotion. However, aggregation procedure of these emerged might be...
An automatic vision-based sewer inspection plays a key role of sewage system in modern city. Recent advances focus on utilizing deep learning model to realize the system, benefiting from capability data-driven feature representation. However, inherent uncertainty defects is ignored, resulting missed detection serious unknown defect categories. In this paper, we propose trustworthy multi-label classification (TMSDC) method, which can quantify prediction via evidential learning. Meanwhile,...
Detecting the anomalous information in multimedia is valuable to many computer vision applications. Recently, pixel-wise methods modeling by deep learning model have been presented, which can be divided reconstruction-based and distance-based methods. However, suffer from low precision of pixel reconstructions. Distance-based extract hierarchical features a pre-trained model, order estimate anomalies distances between normal features. Nevertheless, multi-level are ignored these methods,...
Chaotic time series prediction has attracted much attention in recent years because of its important applications, such as security analysis for random number generators and chaos synchronization private communications. Herein, we propose a BLSTM convolution self-attention network model to predict the optical chaos. We validate model's capability direct recursive prediction, dramatically reduces accumulation errors. Moreover, duration is increased with comparative accuracy where predicted...