Driver Assistance System Based on Deep Learning for Libyan Road Environment

As traffic density continues to increase globally, the frequency of road accidents is expected to rise correspondingly, necessitating the development of driver assistance systems. This paper proposes a dual-camera system to monitor driver drowsiness and detect objects and traffic signs along the driving path. The system integrates two distinct models for drowsiness detection and object recognition. The drowsiness detection model utilises a driver-facing camera, employing a Haar Cascade classifier to identify facial features and monitor eye states. When the system detects that the driver’s eyes remain closed for more than one second, an alert is issued to prompt the driver to regain attentiveness. The YOLOv5-based object detection model processes video frames from a road-facing camera to identify relevant objects and traffic signs in real time. The drowsiness detection model was trained locally and deployed on a laptop for real-time testing in Libyan road conditions, whereas the object detection model was trained on a cloud-based server using Google Colab. The findings of this research indicate that the proposed system has significant potential to enhance road safety through real-time drowsiness monitoring and object detection in the Libyan road environment.