Reliable Drowsy Driver Detection Using Machine Vision and Deep Learning with Rich Facial Feature Analysis

Drowsy driving is one of the major causes of car accidents leading to a huge number of injuries and deaths every year. This paper presents a comprehensive study about developing an AI-based system to detect drowsiness consisting of three sub-systems which are eye's state detection, yawn detection, and head tilde detection. The first approach is using CNN and an augmented image dataset to detect the eye's state and yawning where an accuracy of around 91% was achieved for eye detection and around 75% for yawn detection. The other technique used was using face mesh which extracts facial landmarks and identifies 468 points in the face. MAR and EAR can be calculated using these landmarks. In addition, the head tilde angle can also be calculated using these landmarks. Finally, all the sub-systems will be integrated into one whole system with a GUI for easieruse by users and multi-threaded to run in parallel and avoid single point of failure.