Deep Learning Model Enabled for Distracted Driver Behavior Detection

Abstract

This research presents a novel method for detecting distracted driver behaviour by utilising Deep Learning (DL) models.  The proposed architecture comprises three deep learning models: AlexNet, VGGNet and ResNet. These models utilise a state farm dataset that includes data on one safe driving class and four risky behaviours namely Talking on the phone, Texting on the phone, Turning and Other activities. Distracted driving remains a major contributor to road accidents worldwide. However automated detection technologies show great potential in improving road safety. The experimental results indicate that AlexNet and VGGNet achieved accuracy rates of 98% as well as 98.4% respectively showcasing their efficacy in detecting distracted driving behaviours. This study highlights the significance of choosing deep learning architectures that are customised for certain objectives. Our proposed models have been compared to existing state-of-the-art works and the study confirms that our models perform competitively. The results add to the progress of automated systems designed to decrease accidents resulting from distracted driving highlighting the potential of deep learning in enhancing road safety.