Analysis and Research on YOLOv5s Vehicle Detection with CA and BiFPN Fusion

Abstract

An algorithm based on improved YOLOv5s is proposed to solve the problems of false and missing vehicle detections. Firstly, a coordinate-attention (CA) module is added to the backbone feature of an extraction network to obtain more important information during feature extraction and improve object detection accuracy. Then, the weighted bi-directional feature pyramid network (BiFPN) is adopted to replace the original PANet structure in the YOLOv5s network. This method enhances the multi-scale feature fusion of the model and improves the fusion efficiency. Experiment results present that the mean average precision (mAP) of the improved YOLOv5s algorithm on the BIT-Vehicle Dataset reaches 94.S%, which is 2.S% higher than that of the original YOLOv5s network, and the processing frame rate reaches 136.9, which allows real-time detection by satisfying its requirements.