Analysis and Implementation of Deep Convolutional Neural Network Models for Intelligent Driver Drowsiness Detection System

Driver drowsiness is one of the leading causes of road accidents. An intelligent system to detect and warn drowsy driver is the need of the day. So in this paper design and implementation of an Intelligent Drowsiness Detection System (IDDS), using Deep Convolutional Neural Network (DCNN) model based approach is discussed. Here the effectiveness of 3 DCNN models mainly VGG16, InceptionV3, and Xception is analysed for detecting the eye features of a drowsy driver. A collection of video recordings in both drowsy and awake stages is used for the analysis. The eye closure feature is taken from the videos and used as input to the DCNN models for classification. A comparative analysis of various algorithms reveals that Xception model is best suited for driver drowsiness detection as it gives the highest accuracy. Hence the effectiveness of Xception model is tested using hardware and results are verified.