Pinned Loading

1. Face_Recognition Face_Recognition Public

   Built and deployed a real-time facial-recognition attendance system that automated employee check-in/check-out and achieved 97% accuracy.

   PureBasic 2

2. CCTV-Video-Filtering CCTV-Video-Filtering Public

   Built a VLM-based CCTV video classification module using Qwen3-VL and InternVL3, achieving 94.44% accuracy on a labeled test set and supporting a live demo.

3. Density-Aware-Adaptive-Refinement-Technique Density-Aware-Adaptive-Refinement-Technique Public

   Developed a training-free GUI grounding refinement method that improved KV-Ground-8B accuracy from 73.0% to 80.9% on ScreenSpot-Pro.

4. Restaurant-Traffic-Forecasting Restaurant-Traffic-Forecasting Public

   Improved a forecasting pipeline for 440+ restaurants, reducing WAPE from 25.3% to 22.0%, feature count by 90%, and training time by 70%.

   Python

5. Electricity-Consumption-Forecasting Electricity-Consumption-Forecasting Public

   Developed monthly electricity-demand forecasting models that achieved below 6% MAPE across 13 provinces and below 4% MAPE in six provinces.

   Jupyter Notebook

6. Vision-based-Communication Vision-based-Communication Public

   Implemented a YOLOv11 and ByteTrack-based vision pipeline that improved beamforming data rate by 20–29% on the DeepSense 6G dataset.