RTD3 – Real Time Driver Drowsiness Detection System Using CNN on Inception V3 Model

Abstract

This paper suggests a behaviourally based driver sleepiness detection system. Faces provide data that may be used to gauge sleepiness levels. The level of tiredness may be determined by extrapolating a variety of facial functions from the face. They include head movements, yawning, and eye blinks. Yet, since it needs accurate and reliable algorithms, creating a sleepiness detection system that produces accurate and trustworthy results is a difficult issue. In the past, a wide range of methods have been explored to detect driver sleepiness. The most recent growth in deep learning requires that those algorithms be reviewed to evaluate their efficacy in detecting sleepiness. In this study, the eyes the body feature that matters the most—are used to identify driver sleepiness. The algorithm has been taught to examine the face, isolate the ROI (Region of Interest), or the eyes, and determine whether they are open or closed. When the eyelids are closed continuously for a certain amount of time, the alarm begins to beep to warn the driver. CNN serves as our classifier (Convolution Neural Networks). After a comprehensive review of the literature, we conclude that, in order to get the best outcomes, CNN is the best method to use when compared to vector system approaches, Markov models, etc. The InceptionV3 model, which is a trained image recognition model, is the one employed in this instance. Also, we go through the installation, training, and testing accuracy methodologies.