Reducing Road Mishaps: A CNN Model for Driver Fatigue Detection

Abstract

Over the last few decades, there have been substantial developments in a variety of domains, including computer science, artificial intelligence, and machine learning, which has accelerated the evolution and application of intelligent systems in specific areas such as transportation. One way these systems are used in transportation is through fatigue detection, and to enable this technology, Convolutional Neural Networks serve as the foundation on which this system is built. CNNs are frequently employed in computer vision applications because they can automatically extract pertinent features from input data without the need for manual feature engineering. This CNN-based fatigue detection system was built in-order to address the repercussions of fatigued driving by monitoring the driver’s facial features in real-time so as to predict the fatigue levels and send out an alert to restore the driver’s alertness level so as to reduce road mishap. The system relies on visual input from a camera, which sends this input to a fatigue detection program installed on a Raspberry Pi microprocessor and sends out via alerts via a display and audio alerts via a piezoelectric buzzer. The model shows excellent performance, consistently achieving high accuracy, peaking at 97.34% on the 64th epoch, and consistently maintaining high validation accuracy, reaching 96.02% in the 67th epoch, indicating its ability to function effectively.