Real Time Prevention of Driver Fatigue Using Deep Learning and MediaPipe

Abstract

 This paper describes the development of a  system for detecting driver drowsiness whose goal is to alert  drivers of their sleepy state to prevent traffic accidents. It is  essential that drowsiness detection in a driving environment  be conducted in a non-intrusive manner and that the driver  not be troubled by alerts when they are not sleepy. We make  use of the MediaPipe Facemesh framework to extract facial  features and the Binary Classification Neural Network to  precisely detect drowsy states in our solution to this open  problem. The solution that minimize false positives is created  to determine whether or not the driver exhibits sleepiness  symptoms. The approach extracts numerical features from  images using deep learning techniques, which are then added  to a fuzzy logic-based system. This system typically achieve  91% accuracy on training data and 92% accuracy on test  data. The fuzzy logic-based approach, however, stands out  because it doesn't raise erroneous alerts (percentage of  correctly identified footage where the driver is not tired).  Although the findings are not particularly satisfying, the  recommendations offered in this study are promising and  may be used as a strong platform for future work.