Real-Time Driver Drowsiness Detection Using Transfer Learning

Among the primary causes of traffic accidents, drowsy driving is identified, which poses a significant threat to road safety on a global scale, emphasizing the need to address this issue effectively to ensure road safety. The consequences of drowsy driving are far-reaching, impacting numerous lives and underscoring the urgent need for a real-time system capable of early and accurate detection of driver sleepiness. Addressing this critical issue, this research introduces a machine learning model to monitor driver drowsiness and classify the drowsiness status. Leveraging the advancements in Transfer Learning techniques and utilizing the ResNet50 model specifically trained for drowsiness detection. Through extensive evaluations conducted using the NTHU-DDD dataset, the proposed machine learning model has consistently demonstrated superior performance compared to recent advanced approaches in drowsiness detection. The achieved validation accuracy of 87.8% with the optimized ResNet50 model highlights the system's reliability and potential to impact road safety significantly.