Improved YOLOv8n for Concealed Object Detection in Human Body Millimeter Wave Images
Jinxin Dou, Weixian Tan, Pingping Huang, Yun Su, Yuxin He, Jianxin Zhang, Yanmin Chen, Zhenkun Shen, Ze Li · 2024
This paper, which builds upon the fundamental YOLOv8n architecture, adds the CAFM module to the neck network with the goal of improving detection accuracy and increasing sensitivity of the model to objects in millimeter wave images. This addresses the issue of insufficient detection in millimeter wave images that arises from the difficulty of extracting features of concealed objects. Subsequently, Inner-CIoU is employed as the loss function for bounding box regression, which aims to speed up the convergence of the model. According to experiments, the enhanced YOLOv8n model considerably raises mAP when compared to the baseline model.