Research on Pedestrian Detection Algorithm Based on Improved YOLOv5

Abstract

In response to the random distribution and dynamic characteristics of pedestrians, as well as the false alarms and omissions in the pedestrian detection process, this paper proposes an improved pedestrian detection algorithm based on YOLOv5s. The proposed algorithm incorporates an attention mechanism module by replacing the four C3 modules in the Backbone with CBAMC3. This is done to enhance the network's ability to extract pedestrian target features. The Concat module is replaced by BiFPN to optimize the performance of object detection. Experimental results show that the improved algorithm in this paper has improved Precision, Recall, and mAP compared to directly using YOLOv5s. This indicates that the improved algorithm has significant enhancement effects and can meet the requirements for pedestrian detection, thus having important application value.