Vehicle Detection Based on Improved Yolov5s Algorithm

Vehicle detection technology has been widely used in the field of intelligent transportation, and the performance of existing vehicle detection technology in both detection accuracy and detection speed has been continuously improved. However, when encountering complex road environments, problems such as low vehicle detection rate and poor real-time performance can occur. To address these problems, an improved YOLOv5s vehicle detection algorithm is proposed. Firstly, in the feature fusion module of neck part, a new detection scale is added and the original FPN+PAN structure is replaced with an improved Bi-directional Feature Pyramid Network (BiFPN). Secondly, the Triplet Attention (TA) module l is added to the backbone part and the improved neck part to enhance the feature extraction capability. Finally, the improved algorithm is tested on the MS COCO 2017 dataset, and the experimental results show that the algorithm improves the mean average precision (mAP) by 1.34% to 67.64% compared with the original YOLOv5s algorithm. The detection effect of small-scale vehicle targets is better than the original YOLOv5s algorithm, and the detection accuracy is higher.