DrowsyDetect: Android Application for Driver Drowsiness Detection Using Eye Closure Rate with Deep Learning

Abstract

Studies on sleep-related crashes show that about 15% of accidents are classified as sleep-related accidents. Furthermore, daytime sleepiness is common in young drivers. This means the chances of car crashes due to drowsiness are high. The proposed research attempts to tackle this issue and alert the driver in case of drowsiness and bring the driver back to the consciousness of the surroundings to avoid crashes. The current study aims to develop an Android mobile application, which will be used for drowsiness detection in drivers. It compares multiple Deep Learning, Convolutional Neural Network (CNN) models including MobileNet, VGG16, and a custom CNN to detect user’s faces and analyze the eye closure rate. The models use Haar Cascade to detect the human face and eyes. If the closure rate goes above a certain threshold, an alarm will be triggered on the mobile phone alerting the driver of their drowsiness. The trigger will then be logged on to the user account to allow users to access the logs later and analyze their drowsiness patterns also. The proposed system is simplified to eliminate the need for external equipment like head mounts or sensors that may cause driver discomfort.  However, there is scope for enhancing the app by adding complex features such as sleep schedules, sleep-related information, facts, emergency triggers, and so on.