Driver Assistance System: Utilising Machine Learning for Reducing Accidents, Vehicle and Road Safety

Abstract: The road safety is an important aspect in the present scenario. The project aims to improve the road safety by using machine and deep learning to monitor and classify driver behaviour in real time. It identifies ten types of activities of driver including- safe driving, texting, phone usage, drinking and more. The system uses advances CNNs, transfer learning models like VGG16 and ResNet50, and YOLOv8 for object detection. It also includes a drowsiness detection module to alert drivers showing signs of fatigue. The project uses the state farm distracted driver detection dataset for training and evaluation, and flask-based web app for real-time monitoring and alerts. Performance is measured using Accuracy, Precision, Recall, and F1-score, showing high effectiveness in enhancing driver awareness and reducing accidents. This system is suitable for modern vehicle safety and fleet management solutions. The drowsiness module is also integrated to alert the driver feeling drowsy and improve the safety. It utilizes the standard dataset of open and closed eyes for training and detects the drowsy behaviour in real time.

Keywords: ML, Road safety, VGG16, ResNet50, YOLOv8, CNN, Real-time monitoring.