A5082427311- Fire Detection and Safety Systems
- Video Surveillance and Tracking Methods
- Imbalanced Data Classification Techniques
- Image and Signal Denoising Methods
- Chaos-based Image/Signal Encryption
- Advanced Data Compression Techniques
- Cybercrime and Law Enforcement Studies
- Asphalt Pavement Performance Evaluation
- Infrastructure Maintenance and Monitoring
- Advanced Neural Network Applications
- Advanced Graph Neural Networks
- Data Mining Algorithms and Applications
- IoT-based Smart Home Systems
- Electricity Theft Detection Techniques
- Vehicle License Plate Recognition
Beijing University of Technology
2023-2024
Southwestern University of Finance and Economics
2024
North China University of Technology
2024
Node classification in graph learning faces significant challenges due to imbalanced data, particularly for under-represented samples from minority classes. To address this issue, existing methods often rely on synthetic over-sampling techniques, introducing additional complexity during model training. In light of the faced, we introduce GraphECC, an innovative approach that addresses numerical anomalies large-scale datasets by supplanting traditional CE loss function with Enhanced...
AbstractThis paper describes a scheme for extracting coefficient values from the region of interest (ROI) an image in JPEG2000. Three copies extracted data were hidden three different complex processes non-ROI (NROI) frequency domain before entropy coding stage. This is robust. Experimental results showed that tamper can be detected and tampered areas identified. The original recovered tampering with high peak signal-to-noise ratio no visually detectable distortions. powerful detection...
Imbalanced datasets, where the minority class is underrepresented, pose significant challenges for node classification in graph learning. Traditional methods often address this issue through synthetic oversampling techniques class, which can complicate training process. To these challenges, we introduce a novel paradigm on imbalanced graphs, based mixed entropy minimization (ME). Our proposed method, GraphME, offers 'free imbalance defense' against without requiring additional steps to...
Pavement disease identification based on target detection and image segmentation is a common method in the maintenance of pavement diseases, which has great study value for repair road health status assessment. The use as target, building segmention data set contains pothole, lateral crack, longitudinal crack mesh crack. For task, model improved YOLOv5, model's mAP 94.2%. U-Net, mIOU 94.4%.