This project implements a hybrid computer vision system that analyzes driving behavior using a live camera feed. The system performs real time vehicle detection and tracking, extracts motion-based features, infers driving behavior, and generates a post drive driver skill report.
- Real time vehicle detection using YOLOv8
- Multi object tracking with persistent vehicle IDs
- Motion feature extraction (speed, acceleration, lane stability)
- Rule based driving behavior analysis
- Risk aware real-time visualization
- Post drive driver skill scoring and report generation
- Camera only solution (no vehicle sensors required)
Camera / Video Feed
→ Object Detection (YOLOv8)
→ Object Tracking
→ Motion Feature Extraction
→ Behavior Analysis
→ Live Monitoring + Data Logging
→ Post-Drive Driver Report
Run with live camera:
python src/main.pyRun with video file:
python src/main.py --source data/input_videos/sample.mp4- Real time annotated video feed
- Logged driving behavior data
- Post drive driver skill report
- Fleet driver performance evaluation
- Driving school feedback systems
- Smart city traffic behavior analysis
- Insurance risk assessment research
- Relative speed estimation (no real world calibration)
- Rule based behavior inference
- Camera angle dependency
- Speed calibration using camera geometry
- ML based behavior classification
- Multi camera fusion