Driver distraction and drowsiness continue to be the main cause of road traffic accidents, endangering drivers and pedestrians alike. This research has addressed these issues by proposing a real-time system capable of identifying distraction and drowsiness utilizing deep learning techniques. It combined YOLO (You Only Look Once) object detection technique, to detect driver actions (e.g., phone use, eating) and a Convolutional Neural Network (CNN), for analyzing eye behaviours from tracking that produces drowsiness based on eyelid movement, which are also monitored with YOLO. This project is not meant to replace a driver's real-time actions, but rather to identify potential distractions or drowsiness across a range of detection layers the researcher has detected with YOLO CNN. This paper uses camera input (rather than physical sensors or high end devices) with intelligent algorithms to administrate a structured functional system of potentially dangerous actions/errors involving a driver. Additionally, facial recognition is also used to analyse disturbances in actions and behaviours and provide a security level to vehicle access and detect unauthorized users in real-time. The proposed system is lightweight, can be utilized in different lighting conditions and can provide high correctness with minimum latency, it is applicable to driver assistance systems, combining detection layers and providing a couple layer of driving responsibility to enhance safety in the driving environment.
Faisal et al. (Tue,) studied this question.