DrowsiGuard: Machine Learning-Based Driver Drowsiness Detection System for Enhanced Road Safety

Drivers need to stay focused on the road to respond quickly to unexpected situations. Fatigue is a significant factor in many traffic accidents. Therefore, it's important to have measures that can identify and alert drivers to their deteriorating mental and physical state, potentially reducing fatigue-related incidents. This issue is becoming more prevalent in modern times, leading to an increase in accidents. Various facial expressions and physical signs, like tired eyes and yawning, are acknowledged as indicators of driver fatigue and drowsiness. Using vision-based techniques is one of the technical solutions for creating driver drowsiness detection systems. We introduced a classifier for detecting driver drowsiness using Convolutional Neural Networks (CNN). We conducted experiments using four different CNN architectures: RESNET-34, Kinetics 400, CSPDarknet 53, and DenseNet, and discussed the results. The experimental findings of these models in a driver drowsiness detection classifier represent a comprehensive approach to improving road safety. Among the models tested, ResNet-34 showed superior accuracy and AUC-ROC, offering effective solutions for preventing accidents related to drowsiness.