Vehicle detection algorithm based on lightweight YOLOX

Nowadays, accurate and fast vehicle detection technology is of great significance for constructing intelligent transportation systems in the context era big data. This paper proposes an improved lightweight YOLOX real-time algorithm. Compared with original network, speed accuracy new algorithm have been fewer parameters. First, referring to GhostNet, we make a design backbone extraction which significantly reduces network parameters, training cost, inference time. Furthermore, by introducing α-CIoU loss function, regression bounding box (bbox) improved, while convergence model also accelerated. The experimental results show that mAP on BIT-Vehicle dataset can reach up 99.21% 41.2% parameters 12.7% higher FPS than demonstrate effectiveness our proposed method.