Machine Learning and Artificial Intelligence Techniques for Detecting Driver Drowsiness

Abstract

The number of vehicles on the road rises in tandem with the development of vehicle manufacture. With the rapid growth of automobiles, the number of road accidents appears to be steadily increasing. Accidents are a typical occurrence in our daily lives. Road accidents are the top cause of death from injuries worldwide in the top 10. It now forms an amazingly important part of the global burden of illness. An estimated 1.2 million people are killed in accidents every year, and an estimated 50 million are injured while receiving bone marrow transplants in leading countries. Drowsiness, along with fatigue of drivers, are some of the major causes of road accidents. Every year, we see an increase in death numbers worldwide. This study also looks into automatically detecting driver drowsiness with AI, visuals, and CNN. The Driver Drowsiness System works on the concept of Multi-Layer Feed Forward Network, especially on Convolutional Neural Networks (CNN). CNN was developed with the help of around 7000 images of eyes in both drowsiness and non-drowsiness states with different facial architectures. These images were split into training (80 percent of images) and testing (20 percent of images) datasets. The images under the training dataset are given as the input for the network for training purposes. Backpropagation algorithms and optimizers are used to reduce the loss as much as possible. We created an algorithm for detecting, tracking, and analysing motor and visual effects to measure ROI.