Human Driver Drowsiness Detection

Car accidents caused due to the driver being distracted or falling asleep are on the rise partially due to the advancement in technology and our irregular sleeping schedules. This project attempts to address this issue by creating a driver drowsiness detection and alert system to help drivers maintain focus on the road while driving. We introduce a detection system built entirely as a mobile application that a driver can simply install on his/her iphone.

We make use of recent advancements in Computer Vision, Artificial Intelligence and processing capabilities available in modern phone to make a real-time efficient detection and alert system. The entire system runs at near 30 FPS so as to ensure that the drowsiness detection is fast and prompt alerts are sent to the user to prevent any impending accidents. We also introduce a novel driver sobriety test to verify if a driver is sober before start of a specific drive. Research efforts were focused on creating a low inference cost Deep Learning architecture which would help locate a driver’s face. Parallely, another architecture was designed which would determine if the located face was looking drowsy or not.