Drowsy Driving Detection Based on Deep Neural Network for Accident Avoidance

Abstract

AbstractRoad accidents can be very fatal at times and may lead to loss of lives. Driver drowsiness, a prime cause of road accidents, increases fatality and amount of deaths every year globally. Thus, drowsiness detection among drivers plays a major role in prevention of sleep, thereby reducing road accidents and injuries. In this paper, a deep learning-based architecture for drowsiness detection is proposed. The base architecture used for training and testing is Convolutional Neural Network (CNN) wherein two different CNN architectures, viz., InceptionResNetv2 and ResNet152v2 are used yielding 99.87% and 99.99% accuracy, respectively. For detection purpose, Faster Region-based Convolutional Neural Network (F-RCNN) is applied. The proposed method would be beneficial in terms of safety measures for developing automated monitoring system which can detect driver’s drowsiness instantly.KeywordsDrowsiness detectionDeep learningConvolutional neural networks (CNN)Faster region-based convolutional neural network (F-RCNN)