Detecting Drowsy Driving Through Eye Closure and Yawning : Azure Custom Vision System

With growing concerns about road safety, the increasing number of accidents caused by driver drowsiness has become a major issue. This study addresses the detection of drowsy driving by employing deep learning techniques to recognize sleep states in drivers. Azure Custom Vision model is propounded to identify closed eyes in constant facial images of drivers. The proposed approach offers versatile applications such as human-computer interaction design, facial expression recognition, and fatigue detection in drivers. The AZURE Custom Vision utilizes the Convolutional Neural Network VGG16 model with Azure. A dataset is created by capturing the drowsy driving instances and are trained. Our model achieves a total detection accuracy of 99.3%, outperforming other existing algorithms in the field. For instance, RNN achieved an accuracy of 92%, while CNNs reported an accuracy of 96.51%. Compared to these models, our approach demonstrates superior accuracy, leveraging the deep hierarchical feature extraction capabilities of VGG16 to more effectively distinguish between drowsy and wakeful states.