摘要
传统人脸检测网络在复杂背景且人脸有偏转角度情况下一直存在检测效率低且检测准确率低等问题。针对以上问题,论文对多任务卷积神经网络(MTCNN)进行了改进,改进后的网络改变原始论文进行非极大值抑制的方法,并且引入SE Module注意力机制。将改进后的网络在AFLW人脸公开数据集上进行实验。试验结果表明改进后的算法可以在原始MTCNN算法的基础上进一步提高在人脸有偏转角度情况下的检测鲁棒性,并且检测准确度和检测速率都有所提升。
The traditional face detection network always has the problems of low detection efficiency and low detection accuracy in the case of complex background and face deflection angle.To solve the above problems,this paper improves the multitask convolutional neural network(MTCNN).The improved network changes the non-maximum suppression method of the original paper,and introduces the SE Module attention mechanism.The improved network is tested on AFLW face public data set.Experimental results show that the improved algorithm can further improve the detection robustness when the face has deflection angle on the basis of the original MTCNN algorithm,and the detection accuracy and detection rate have been improved.
作者
杨玉洁
陈天星
石林坤
袁标
YANG Yujie;CHEN Tianxing;SHI Linkun;YUAN Biao(College of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031)
出处
《计算机与数字工程》
2023年第9期2074-2078,2206,共6页
Computer & Digital Engineering
