参考文献-华东理工大学学报.DOC

__________________________收稿日期：基金项目： 作者简介：赵澜涛(1990-)，男，湖北人，研究生，主要研究方向为机器学习和多姿态人脸识别通讯联系人：林家骏，E-mail: jjlin@ecust.edu.cn. 基于双路CNN的多姿态人脸识别方法 赵澜涛， 林家骏 (华东理工大学 信息科学与工程学院，上海200237) 摘 要：本文提出一种基于CNN（Convolutional Neural Network，CNN）的多姿态人脸识别方法。利用该方法能将输入的人脸投影到高维特征空间并输出具备姿态鲁棒性的人脸特征，从而进行精确的多姿态人脸识别。经过大量实验，该模型在多个数据集上取得良好的效果。与传统的单路CNN网络层次结构不同，本文中采用双路CNN网络层次结构并结合度量学习来优化传统的CNN模型。最后，本文使用tensorflow深度学习框架进行实验，实验结果表明该框架的识别准确率比当前几种常用的多姿态人脸识别算法要高。 关键词：多姿态人脸识别；卷积神经网络；深度学习 中图分类号：TP391 文献标志码：A Pose-invariant Face Recognition with Two-Pathway Convolutional Neural Network Lan-tao ZHAO, Jia-jun LIN (School of Information Science and Engineering,East China University of Science and Technology,ShangHai 200237 ,China) Abstract：Face recognition has always been a hot topic in the field of computer vision and pattern recognition. With the development of deep learning and the improvement of computer computing performance, a fairly good recognition rate has been obtained on multiple datasets. But the high recognition rates are only exist in the input image sets are obtained by a standard frontal pose, when face pose changes the recognition rate will decline rapidly, especially in the case that the face angle at 45 degrees to 90 degrees .This paper proposes a valid method for pose-invariant face recognition based on Convolutional Neural Network(CNN). The network can project the input face to high dimensional feature space ,explicitly disentangle identity and pose information in the latent feature space. We obtained good experimental results on several face datasets. Different from the conventional deep learning methods that mainly rely on single path CNN structure, our method rely on the two-pathway CNN structure and metric learning to obtain a feature representation that is invariant to fase pose. In the image preprocessing stage, we make use of the ready-made dlib C++ Library to detect the face and wipe out of the effect of the background. During the Model phase, we design a