广西科技大学学报2025，Vol.36Issue(5)：52-57,64,7.DOI:10.16375/j.cnki.cn45-1395/t.2025.05.007
基于YOLOv5的校园行人及人脸检测研究
Campus pedestrian and face detection based on YOLOv5
摘要
Abstract
Pedestrian and face detection on campus is fundamental to intelligent campus security.However,existing deep learning detection methods are not specifically designed for the large number of multi-scale pedestrian objects on campus,and there are few models that can achieve a balance between performance and the number of parameters in campus management systems with limited computational resources.Therefore,this study designed a low-parameter,high-performance pedestrian and face detection network for campus environments,named Campus-YOLO.In the backbone network,a lightweight module was set up,and the network depth was reduced to decrease the number of network parameters and improve detection accuracy,making it more suitable for campus management systems with limited computational resources.Moreover,a novel attention mechanism was designed to more effectively capture feature information at different scales.In the head part of the model,an Anchor-free strategy was adopted to simplify the model structure and improve training efficiency.Ultimately,with only 1.70 MB parameters,this model achieved an excellent detection performance with a mAP of 0.792,fully meeting the requirements of high detection accuracy and limited computational resources in campus environments.The results of this study provide effective technical support and new insights for intelligent detection in campus security.关键词
校园安全/目标检测/轻量化/注意力机制/多尺度特征Key words
campus security/object detection/lightweight/attention mechanism/multi-scale features分类
计算机与自动化引用本文复制引用
杨浩,林川..基于YOLOv5的校园行人及人脸检测研究[J].广西科技大学学报,2025,36(5):52-57,64,7.基金项目
国家自然科学基金项目(62266006,61866002) （62266006,61866002）
广西自然科学基金项目(2020GXNSFDA297006,2018GXNSFAA138122,2015GXNSFAA139293)资资 （2020GXNSFDA297006,2018GXNSFAA138122,2015GXNSFAA139293）
