Abstract

A growing number of vulnerable road users (VRUs), including cyclists, motorcyclists, and pedestrians, are involved in traffic accidents; therefore, reliable and effective detection technologies are needed to improve road safety. This paper proposes a computer vision approach using You Look Only Once version 8 (YOLOv8) for VRU detection. We adopted the method to detect VRUs by training them on a large dataset of annotated images containing road users with diverse mobility and evaluated it on a custom dataset. We achieved an average precision of 0.85 and an average recall of 0.78 over a significantly lower number of training epochs. The suggested approach can be implemented in real-time applications such as enhanced driver assistance systems, driverless cars, and surveillance systems to enhance VRU identification. This research contributes to the ongoing efforts to create safer road environments by leveraging the capabilities of modern computer vision and deep learning technologies. Integrating YOLOv8 and an onboard vision-based approach holds significant promise for reducing accidents involving vulnerable road users and fostering safer coexistence between vehicles and VRUs on today’s busy roadways.