Safety on Wheels: Computer Vision for Driver and Passengers Monitoring

Monitoring passengers and their actions inside a vehicle is one of the most important issues for road and human safety in the whole world. However, the problem revolves around road accidents due to the lack of comprehensive in-vehicle monitoring systems, and the potential hazards including driver distraction, drowsiness, violence, and emotional distress. This research presents a comprehensive framework that utilizes deep learning techniques to enhance passenger and driver safety. We used the YOLOv8 model to address each seat-belt detection, drowsiness detection, and driver distraction detection which achieved testing accuracy of 78 %, 65 %, and 70 % respectively. Driver verification was accomplished using the DeepFace library with a testing accuracy of 100 %. Our system also performs emotion detection using a fine-tuned CNN model for face recog-nition which achieves an accuracy of 85.5 %. Additionally, we perform violence detection using a VGG19-LSTM which achieves an accuracy of 100%.