Detection of Driver Drowsiness for Road Safety using Convolutional Neural Network

Driver drowsiness is one of the major reasons for road accidents worldwide.In this paper, a real-time driver drowsiness detection system based on Eye Aspect Ratio (EAR) estimation using Convolutional Neural Network is presented.The system tracks the pattern of eye closure from the video frames captured by an inbuilt webcam, estimates EAR from facial landmarks, which is then passed to a CNN that separates the open and closed eyes.The system alarms on detecting drowsiness, with the aim of avoiding fatigue-related accidents.The suggested technique is 97.2% accurate on the fused dataset and works efficiently in real time, even under changing light and user conditions.The research can be the building block to integration in contemporary transportation systems like ridesharing and logistics.