Vehicle Target Detection with an Integrated Attention Mechanism

To address the problems of traditional vehicle detection models regarding equipment requirements, detection accuracy, and overlapping target leakage rate, this study proposes an improved vehicle target detection algorithm for YOLOv8. Firstly, we introduce EMA, an attention mechanism for cross-space learning, to improve the attention to feature information. Secondly, it combines EMA with the C2f module in the Neck network to fuse more important feature information, obtaining more accurate vehicle feature information. Finally, it uses the MPDloU loss function to speed up the model's convergence and improve detection accuracy. Experiments conducted on Pascal VOC and MS COCO datasets show that the performance of the improved algorithm proposed in this paper exceeds that of the original model. The enhanced model's mAP reaches 85.71%, 3.27% higher than the original model, improving vehicle target detection accuracy and being highly practical.