基于混合训练的深度学习人脸特征提取方法

摘要

The face recognition method based on deep learning is studied.In terms of the various methods of training network for face feature extraction,the advantages and disadvantages are analyzed and compared.The characteristics of the comprehensive application of various methods is put forward to extract the depth of facial feature of mixed training,and the extraction method is used to train facial features a mixture of Softmax and Triplet loss.The depth of the facial feature ex-traction method of mixed training has obvious advantages in terms of accuracy and the training speed,and the same out-standing performance on face authentication,face recognition and face retrieval tasks.The method is helpful for improving the efficiency of military management and realizing the modernization of military management.%通过对基于深度学习的人脸识别方法进行研究,对人脸特征提取中进行训练网络的各种方法的优缺点进行分析和比较,提出了一种基于混合训练的深度学习人脸特征提取方法,即采用混合Soft-max和Triplet函数训练的人脸特征提取方法.基于混合训练的深度学习人脸特征提取方法在准确度、训练速度等方面优势明显,在人脸验证、人脸识别和人脸检索任务中同样表现出色.该方法为提升军事管理效率,实现军事管理现代化提供了帮助.