DRIVER DROWSINESS DETECTION USING PYTHON AND MACHINE LEARNING

This work presents a new driver drowsiness detection system implemented with Python and machine learning (ML) methods for improved road safety. The system makes use of the eye aspect ratio (EAR) as the major measure to detect drowsiness, supplemented by yawn detection (through mouth aspect ratio, MAR) and eye focus monitoring (through gaze tracking and blink frequency analysis). The method uses real-time video processing with OpenCV, dlib-based facial landmark detection, and an SVM-trained ML classifier based on extracted features. Experimental testing with 20 volunteer drivers under different conditions (daylight, nighttime, and simulated drowsiness) resulted in a 92% accuracy of drowsy state detection with a false positive rate minimized by 15% through multi-feature integration. The findings highlight the potential for real-world deployment of the system, though areas of improvement include low-light performance (85% accuracy) and computational loads. The research makes a scalable, costeffective contribution to the field of intelligent transportation, with future improvements suggested in the form of larger data sets and hardware integration.