Driver Drowsiness Detection using Convolutional Neural Networks(CNNs)

Abstract

Driver drowsiness is a significant factor in many fatal and serious traffic incidents. We suggest a novel method for detecting driver tiredness using a Convolutional Neural Network (CNN) based on yawn detection and eye blink rate analysis in order to address this problem. In order to extract useful information from facial photos taken by an onboard camera, our method makes use of a CNN architecture. Our method's first step focuses on yawn recognition, in which we use a trained CNN to spot yawning in the driver's facial expression. Yawning is a reliable indication for identifying driver weariness because it is a typical sign of drowsiness and exhaustion. We include eye blink rate analysis as another crucial sign of sleepiness in addition to yawn detection. We use image processing methods to track and count the number of eye blinks over a predetermined period of time. We can determine the driver's level of tiredness by examining the blink rate. A dataset of labelled yawning is used to train the suggested CNN architecture. We employ deep learning methods to acquire discriminative features that allow for the precise detection of driver intoxication. Using data from actual driving, we assess the effectiveness of our system and compare it to other approaches for detecting driver tiredness. Results from experiments show that our method is successful in detecting sleepiness episodes with high accuracy. The suggested method has a great deal of potential for integration with real-time driver support systems, such as mobile applications or in-vehicle monitoring systems, to improve driver safety by offering prompt alerts and interventions in the event of drowsiness detection. Our method provides a solid and dependable solution for driver tiredness detection by utilizing the capabilities of CNNs and integrating yawn recognition with eye blink rate analysis.