Investigation on Driver Drowsiness Detection using Deep Learning Approaches

Driver sleepiness is a major factor in road accidents, necessitating effective detection methods to enhance road safety. Before the emergence of deep learning models, conventional approaches rely on manual observation to identify drowsiness symptoms in drivers. However, these methods are subjective, time-consuming, and prone to errors. This work aims to address the limitations of previous methods and provide a more accurate and efficient solution. To achieve this, a comparative analysis has been conducted using three popular CNN architectures: VGG19, EfficientNetB7, and MobileNetV2. These models are trained and evaluated on a carefully annotated dataset of driver eye images, classifying them as opened or closed. The results demonstrate the effectiveness of deep learning models in accurately detecting driver drowsiness. Among the evaluated architectures, EfficientNetB7 and MobileNetV2 achieve outstanding accuracy rates of 99.87% and 99% respectively, surpassing the traditional approaches. This paper contributes field of driver safety by demonstrating the potential of deep learning models in reducing the risks associated with drowsy driving.