摘要
针对小人脸检测容易出现漏检的问题,对YOLOv5算法进行改进,在YOLOv5的骨干网络中添加通道注意力模块,改进后的算法在Wider Face数据集上的测试识别准确率提高了3%;改进SRGAN算法,用残差密集网络替代生成网络,改进后的算法在LFW数据集上测试,图像质量评价指标PSNR值达到31.5;最后采用FaceNet算法识别人脸。整合以上算法,完成高校课堂考勤系统开发。系统能够对课堂上学生的照片、视频进行人脸识别,并将识别的结果保存到数据库中供用户查看。
To solve the problem of missing detection in small face detection,the YOLOv5 algorithm is improved and a channel attention module is added to the backbone network of YOLOv5.The detection accuracy of the improved algorithm on the Wider Face dataset is increased by 3%.SRGAN algorithm is improved to replace the generated network with the residual dense network.The improved algorithm is tested on the LFW dataset,and the PSNR value of image quality evaluation index reaches 31.5.Finally,FaceNet algorithm is used to recognize faces.Integrate the above algorithms to complete the development of college class attendance system.The system can recognize the faces of students'photos and videos in class,and save the recognition results to the database for users'viewing.
作者
董亚蕾
张师宁
武旭聪
DONG Yalei;ZHANG Shining;WU Xucong(Hebei Chemical&Pharmaceutical College,Shijiazhuang 050026,China;Hebei International Studies University,Shijiazhuang 051132,China)
出处
《现代信息科技》
2023年第12期62-65,共4页
Modern Information Technology
关键词
课堂考勤
小人脸检测
超分辨率重建
人脸识别
class attendance
small face detection
super-resolution reconstruction
face recognition