Driver Distraction Detection

This project uses computer vision to monitor driver behavior through a camera feed. It identifies distracted states such as phone usage, eating, and head-turning using a combination of pose estimation and YOLO-based object detection, then triggers alerts when distraction is detected.

The system is built in Python, leveraging YOLOv11 for object detection and MediaPipe for pose/keypoint estimation with an OpenCV-powered processing pipeline.

I designed and implemented the full detection pipeline — from data collection and model training to building the real-time inference loop and alert system.

• Real-time video stream processing
• Multi-class distraction detection (phone, eating, etc.)
• Pose keypoint analysis for head and hand position
• Visual and audio alert system
• Performance logging for analysis