Driver Monitoring System — Real-Time Attention and Drowsiness Detection

Technologies: Python, OpenCV, MediaPipe (Face Landmarker), NumPy

● Developed a real-time computer vision pipeline, including video capture, facial landmark detection (468-point MediaPipe mesh), and feature extraction, to continuously score driver attention from live camera or video input.

● Computed Eye Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR) per frame to detect eye closure and yawning, and estimated head pose (yaw, pitch, roll via solvePnP) to flag distraction and fatigue in real time.

● Implemented a weighted attention-scoring algorithm combining EAR, head-pose, and yawn signals, paired with an audio and on-screen HUD alert system with cooldown logic to prevent alert fatigue.

● Created a CSV-based session logging module to record attention scores over time, enabling post-session performance analysis.