Anonymous Person Tracking Desktop Application
AnonTrack is a desktop application for anonymous person tracking in video feeds. It detects and monitors individuals' movements while strictly preserving their privacy through real-time face anonymisation — without storing or processing any personally identifiable data.
Built as a BSc thesis project at the Budapest University of Technology and Economics.
- Face Anonymisation — Blur or pixelate faces before any AI inference
- Person Detection — Real-time detection using YOLO11
- Multi-Object Tracking — Consistent anonymous IDs via DeepSORT
- Cross-Video Re-identification — Persistent vector embedding database with cosine similarity matching
- Heatmaps — Visualise crowd density across the video
- Trajectory Maps — Per-person movement paths with direction vectors
- Movement Statistics — Speed, direction, presence %, occupancy and more
- Export — Save heatmaps, trajectory maps, and CSV statistics
| Layer | Technology |
|---|---|
| Language | Python |
| GUI | Qt for Python (PySide6) |
| Face Detection | RetinaFace (InsightFace) |
| Person Detection | YOLO11 (Ultralytics) |
| Tracking | DeepSORT |
| Computer Vision | OpenCV |
| Embeddings | MobileNet + pickle database |
- Upload a video file (.mp4, .avi, .mov, .mkv)
- Select frame interval and anonymisation method
- Run detection and tracking — watch results live
- Explore heatmaps, trajectories, and statistics in the Inference tab
This project is licensed under the MIT License.
Supervisor: Dr. Mohammad Saleem, Budapest University of Technology and Economics