Application of CNN based architectures in detection of distracted drivers

Abstract

Distracted driving is known to be one of the most significant reasons behind the occurrence of traffic accidents. Moreover, the phenomenon of the occurrence of road accidents due to distracted driving has been increasing at a high rate in recent years. Previously, different machine learning and neural network-based approaches were taken to find out the best possible way of detecting distracted driving. This work proposes an effective interpretation which is to detect the distraction of drivers through a Deep Learning approach through the implementation of several Convo lutional Neural Network (CNN) based architectures. The results presented in this research is to confirm the better accuracy and success rate of the Deep Learning approach to detect distracted driving behaviors demonstrating the potentiality of this method to help measure unusual driving performance. The proposed custom CNN model not only ensures an impressive accuracy but also it’s ability to interpret the proper regions of interests on two datasets of distracted driving