Deep Learning Based Continuous Real-Time Driver Fatigue Detection for Embedded System

Traffic accidents are caused by various reasons, including combination of misbehaviors, such as carelessness and negligence, thus, leading to lethal accidents and property loss. Among them, drawsiness is considered as one main reason. As such, we believe a highly accurate, real-time driver monitoring and fatigue detection system can contribute to reduce these accidents. In addition, to be mounted inside the vehicle, such a system should also allow embedded operation. In this study, using Nvidia Jetson Nano, a highly accurate, real-time and lowcost embedded system was propopsed to perform driver fatigue detection and monitoring. Through deep learning based methods, the system classifies four different states using eye and mouth regions of the driver, and determines fatigue status. Experimental investigation reveals encouraging performance of the proposed system