Deep Learning and the Future of Drowsy Driving Prevention

With the number of vehicles in use rising every year, drowsy driving crashes are becoming more prevalent than ever before. This study delves into the application of Artificial Intelligence in risk management; more explicitly, the use of deep learning in computer vision to detect if a driver is alert or in an unsafe position to operate a vehicle. Two AI models were used to train the dataset, those being YOLOv8-m and YOLOv9-gelan-c. YOLOv9-gelan-c is the most recent of the two computer vision models and the most advanced. However, YOLOv8-m was also used to set a benchmark for a consistent and efficient model. After a comparative overview, the YOLOv9-gelan-c model was chosen as the superior model due to its higher average AP scores (90.6% against 88.9%) and the lack of a false positive. These findings are essential for creating an effective drowsy driving prevention system. The model’s performance was also compared against several other studies that used computer vision and detection software. Furthermore, this research promotes the integration of detection software for safety features within the automotive industry, to lower drowsy driving accidents and fatalities.