Driver Drowsiness Detection System

Real-time computer vision system for detecting driver fatigue using CNN and eye-state analysis.

Problem

Driver fatigue is a major cause of road accidents, but detecting drowsiness in real time using lightweight systems remains challenging, especially under varying lighting and head pose conditions.

Solution

Used Haar cascade classifiers for real-time face and eye detection

Trained a CNN model to classify eye states (open vs closed)

Implemented score-based temporal logic to detect sustained eye closure

Integrated audio alert system to notify driver in real time

Tech Stack

OpenCVCNNKerasPygame

Architecture

Webcam Stream → Face Detection → Eye Region Extraction → CNN Classification → Score Tracking → Alert Trigger

Challenges

Ensuring robustness under varying lighting conditions and head movements was difficult. Additionally, reducing false positives while maintaining sensitivity required careful tuning of threshold-based alert logic.

What I’d Improve Next

• Incorporate facial landmark detection for higher accuracy
• Add yawn detection and head pose estimation
• Deploy on edge devices for in-car embedded systems