REAL-TIME DRIVER DROWSINESS DETECTION SYSTEM

Abstract

Driver fatigue and drowsiness pose significant risks to road safety, necessitating advanced monitoring systems to mitigate accidents. This research addresses this concern through the implementation of a Real-Time Driver Drowsiness Monitoring system using Vision Transformer (ViT). The study involved fine-tuning a Vision Transformer (ViT) model with additional layers on a dataset comprising 84900 images of open and closed eyes, enabling accurate detection of driver drowsiness. The motivation for drowsiness detection stems from the critical role of attentiveness in driving safety. Existing alarm systems often rely on conventional methods, and the research seeks to enhance this by incorporating state-of-the-art ViT technology. Achieving an impressive accuracy of 98.8%, the real-time monitoring system uses the ViT model to detect driver drowsiness promptly. When drowsiness is detected, an alarm is triggered to alert the driver, promoting immediate attention, and reducing the risk of accidents. The model's performance is also thoroughly evaluated, demonstrating high precision, recall, and F1-Score for both closed and open-eye states. This study benefits the field of driver safety by providing a reliable and effective real-time drowsiness monitoring method, thus enhancing road safety, and reducing the incidence of fatigue-related accidents

Keywords