A Comparative Study on Procedural and Predictive Systems on Drowsiness Detection

This study addresses the critical issue of drowsiness and driver fatigue, leading causes of fatal road accidents worldwide. The focus is on enhancing transportation safety by implementing an automatic driver drowsiness detection system using Python, incorporating OpenCV, Dlib, Tensorflow, and SciPy. The present study emphasizes the need for technological solutions to continuously monitor and measure driver attention, particularly during extended journeys. The proposed drowsiness detection system identifies closed eyes instances, alerting the driver to prevent potential accidents. A comparison between a deep learning LRCN model and a procedural EAR model shows that the EAR model achieves approximately three times faster execution time and higher accuracy in a GPU-exclusive environment. The procedural model achieves 83-86% accuracy on around 1000 images in 60-67 seconds, outperforming the LRCN model.