Real-Time Driver Drowsiness Detection Using CNN: A Web-Based Application Approach

Driver drowsiness is a significant contributor to traffic accident. Lack of sleep or long driving hours can impair a driver’s attention and mental performance, which raises the possibility of harm or death in accidents. To solve this issue we have introduced Driver Drowsiness Detection System. The fundamental goal of our method is to use computer vision to identify the driver’s level of tiredness based on their eye condition and to trigger an alarm to notify the user when a sleepy state is noted. This study presents a web application based on Convolutional Neural Networks for real-time drowsiness identification with an accuracy of 97%. The methodology involves analyzing facial features, particularly eye closure, using computer vision techniques, ReLU and Softmax activation functions to alert drivers when their eyes remain closed beyond a set threshold. Haar Cascade classifiers are employed for face and eye detection to preprocess image data effectively. The dataset used consists of images of faces and eyes of people sourced from Kaggle. The results shows the efficiency of this approach in real-time detection of drowsiness, achieving 97% weighted average score for precision, recall, and F1 scores across all classes. This system highlights the potential of CNNs and computer vision in addressing the critical issue of driver fatigue. The importance of this research lies in its ability to enhance road safety by developing a robust, real-time drowsiness detection mechanism that can handle safety of drivers.