REAL TIME DRIVER DROWSINESS DETECTION USING TRANSFER LEARNING

Abstract

Driver drowsiness is the result of a driver driving in a tired state. The state of driver drowsiness contributes to accidents on the roads most of the time. Driver drowsiness on the roads might lead drivers to lose their lives or damage properties. Due to the author's concern about this issue, the author developed a system to detect driver drowsiness and prevent the driver from a drowsy state by triggering an alarm. For this project, the author decided to use MobileNet CNN Architecture to classify close or open eye images from MRL Eye Dataset and OpenCV to determine if drivers are drowsy or not by tracking driver eyes' status. A dataset consists of approximately 20,000 images of closed and open eyes randomly selected from the MRL Eye Dataset. Those images were converted to be arrays of data and trained with transfer learning. The purpose of this project is to reduce accidents on the roads due to driver drowsiness. Therefore, lessening road accidents lead to saving many lives on the roads.