Driver Drowsiness Detection Using Deep Learning *

Drowsiness is the main factor that leads to road accidents. Statistics show that drowsy drivers are responsible for accidents. A National Sleep Foundation survey indicates that 20% of drivers are fatigued and sleep deprived when operating a vehicle. This study offers a novel method for detecting driver drowsiness in this setting by utilizing deep learning(DL) techniques. The study leverages the Media Research Lab (MRL) Eye dataset to capture eye states for analysis. To distinguish between drowsy and awake states, single architectures like InceptionNetV3, MobileNetV2, and EfficientNetB2 are used, along with a stacked ensemble model that combines these architectures, achieving highest accuracy of 0.9621%. This study highlights the potential of DL to improve road safety while advancing our awareness of and response to the problems associated with drowsy driving.