Facial Feature-Based Microsleep Detection with High Precision Using Deep Learning

Annually, approximately 1.3 million people lose their lives in road traffic accidents, with microsleep being a significant contributing factor. The rise in microsleep-related accidents is attributed to factors like heavy workloads, extended hours, traffic congestion, and substance consumption. These incidents result in fatalities, injuries, property damage, and disability. To combat this trend, we propose a deep learning-based approach to detect microsleep in smart vehicles. Leveraging 772 images data points from Kaggle, we evaluate our model using the intersection of union (IoU) of microsleep traits. Focusing on facial features, our approach predicts microsleep confidence values, utilizing deep learning's capabilities in image recognition. Initial results are promising, with a mean Average Precision (mAP) of 99%, signifying high accuracy in microsleep detection. This level of accuracy holds the potential to significantly enhance road safety by effectively identifying and preventing microsleep-related accidents