Real-time driver fatigue detection system with deep learning on a low-cost embedded system

Abstract

Road traffic accidents result in significant life and property losses, which are caused by various factors including driver fatigue and drowsiness. Therefore, real-time monitoring of the driver's state inside a vehicle and accurate detection of fatigue is essential to reduce the number of accidents. However, achieving high accuracy with low-cost embedded devices has been a challenge. This study proposes a novel approach that uses deep learning to accurately detect driver fatigue in real-time on the Nvidia Jetson Nano embedded device. The proposed system utilizes deep learning architecture, specifically Convolutional Neural Networks (CNNs), to classify four different situations by analyzing the eye and mouth areas of the driver. In addition, the dlib library is employed to precisely locate the driver's eye and mouth regions. The system is trained and tested on the YawDD dataset and achieves an accuracy of 93.6% and 94.5% for the eye and mouth models, respectively. The system operates at an average speed of 6 fps on the Nvidia Jetson Nano embedded device. The proposed system contributes to the field of driver fatigue detection by addressing the challenges of achieving high accuracy in real-time on a low-cost embedded device. This system aims to minimize the number of accidents and protect human life during transportation by detecting driver fatigue and issuing an alert. The classification results demonstrate the success of the proposed system, which accurately classifies four different states of the driver and detects driver fatigue states with high accuracy. Overall, this study presents a significant contribution to the field of driver fatigue detection by proposing a real-time, low-cost, and accurate system that can be installed in vehicles to ensure safe transportation and prevent accidents.