Improved traffic scene detection algorithm based on YOLOv8

With the rapid advancement of urbanisation, the road traffic flow has increased dramatically, and the traffic scene has become more and more complex, in-depth study of traffic scene target detection technology is of great practical significance to promote the development of intelligent transport system. In this study, an improved version of the traffic scene detection algorithm is introduced for YOLOv8, which introduces the CPMSB structure to enhance the feature extraction capability of the backbone network, and improves the extraction efficiency of small targets in complex environments through multi-scale feature extraction; the necking network is optimised by introducing the multi-branch assisted FPN, which enhances the feature fusion capability and detection performance of the necking network. The experimental results show that compared with the original model, the improved model improves the mAP by 1.1 percentage points and the F1-score by 1.9 percentage points in the road detection task.