Driver Drowsiness Detection Using Machine Learning

Abstract

Driver drowsiness is a critical problem that threatens the safety of drivers operating vehicles, especially drivers who do not take regular breaks when driving long distances, which carries a high risk of death. Among the solutions aimed at saving lives, artificial intelligence, in particular convolutional neural networks (CNN), applied in computer vision applications, have been adapted to predict the state of fatigue and drowsiness of drivers. In this work we used a model that classifies the driver's condition in real time. Based on the training database, YAWDD (YAWNING DETECTION DATASET), is divided into four categories with different postures: closed-eye, open-eye, yawning, and non-yawning to determine the driver's condition. Primacy in the bisection is given to the state of the eyes as an indicator of drowsiness. So, in turn, it is a good indicator to detect the drowsiness of the driver. He can also predict States of fatigue even before drowsiness occurs. We also used several algorithms such as AlexNet, MobileNet, Model c and Comparing their results with all scales, which gave promising results with accuracy (94%, 94% and 87%) respectively which proved that the MobileNet model is more efficient in detecting drowsiness and fatigue