湖北汽车工业学院学报2025，Vol.39Issue(4)：7-12,6.DOI:10.3969/j.issn.1008-5483.2025.04.002
基于改进YOLOv8的密集行人检测算法
Dense Pedestrian Detection Algorithm Based on Improved YOLOv8
摘要
Abstract
In response to issues of missed detections and difficulties in feature extraction caused by overlap,occlusion,and small size in dense pedestrian detection,an improved YOLOv8 algorithm was proposed.In the algorithm,EfficientNet was introduced to optimize the C2f module,enhancing the effi-ciency of feature extraction and feature representation capability.Additionally,the RepGFPN module was introduced to fuse multi-scale feature maps,improving multi-scale detection capabilities.The RFAHead detection head was used to dynamically adjust the receptive field,optimizing target region capture.Experimental results show that on the WiderPerson dataset,mAP50 and mAP50-95 of the im-proved algorithm increase by 2.6%and 3.4%,respectively,with a 28.4%reduction in computational load.On the CrowdHuman dataset,mAP50 and mAP50-95 of the improved algorithm improve by 4.3%and 5.8%,respectively.关键词
密集行人检测/YOLOv8/EfficientNet/RepGFPN模块/检测头Key words
dense pedestrian detection/YOLOv8/EfficientNet/RepGFPN module/detection head分类
信息技术与安全科学引用本文复制引用
Gong Yu,Gao Ren,Xu Longyan,Chen Yaling..基于改进YOLOv8的密集行人检测算法[J].湖北汽车工业学院学报,2025,39(4):7-12,6.基金项目
湖北省教育厅科学技术研究计划项目(D202111802) （D202111802）
湖北省重点研发计划项目(2022BEC008) （2022BEC008）
