福建电脑2024，Vol.40Issue(6)：46-53,8.DOI:10.16707/j.cnki.fjpc.2024.06.008
课堂场景密集人头检测技术
Dense Head Detection Technology in Classroom Scenarios
杨悟琦 1艾旭升1
作者信息
- 1. 苏州工业职业技术学院人工智能学院 江苏 苏州 215104
- 折叠
摘要
Abstract
In the classroom attendance system,head detection is very important for counting the number of people and locating students.To solve the problem of difficult recognition of small targets and occluded objects in densely populated classroom scenes,this paper proposes three classroom head detection models,namely YOLOv5n+GFPN,YOLOv5n+GFPN+C3-Faster,and YOLOv5nl+GFPN+C3-Faster.The loss functions used include CIoU,Focal CioU,and MPDIoU.The experimental results indicate that the MPDIoU loss function combined with YOLOv5nl+GFPN+C3-Faster mAP@50 It is 0.821 and GFLOPs are 7.2,indicating that YOLOv5nl+GFPN+C3-Faster can effectively improve the performance of classroom head detection under the MPDIoU loss function.关键词
目标检测/课堂场景/头部检测模型Key words
Object Detection/Classroom Scenarios/Head Detection Model分类
信息技术与安全科学引用本文复制引用
杨悟琦,艾旭升..课堂场景密集人头检测技术[J].福建电脑,2024,40(6):46-53,8.
