广东工业大学学报 ›› 2018, Vol. 35 ›› Issue (05): 11-19.doi: 10.12052/gdutxb.180031

• 综合研究 • 上一篇    下一篇

基于CycleGAN的非配对人脸图片光照归一化方法

曾碧, 任万灵, 陈云华   

  1. 广东工业大学 计算机学院, 广东 广州 510006
  • 收稿日期:2018-03-08 出版日期:2018-07-10 发布日期:2018-07-10
  • 作者简介:曾碧(1963-),女,教授,博士,主要研究方向为智能机器人、机器学习.
  • 基金资助:
    国家自然科学基金资助项目(61703115);广东省自然科学基金资助项目(2016A030313713);广东省重大科技专项项目(2016B010108004);广东省产学研合作专项项目(2014B090904080)

An Unpaired Face Illumination Normalization Method Based on CycleGAN

Zeng Bi, Ren Wan-ling, Chen Yun-hua   

  1. School of Computers, Guangdong University of Technology, Guangzhou 510006, China
  • Received:2018-03-08 Online:2018-07-10 Published:2018-07-10

摘要: 针对人脸识别过程中光照对识别结果的影响问题,提出了一种基于CycleGAN的光照归一化方法.使用了生成对抗式的网络结构,利用图像翻译的原理,将较亮图片的光照风格迁移至较暗图片,同时保持原人脸表面平滑且结构基本不变.使用非配对的数据集,无需人工标注标签,简化了数据准备阶段的工作,达到了利用无监督的深度学习方法去除图片光照影响的目的.最后用训练好的模型处理CroppedYale测试集,比较处理前后的人脸识别准确率.实验证明,本文方法具有较强的降低人脸光照对识别率影响的能力且基本不改变人脸结构,有利于提高人脸识别的准确率.

关键词: 生成对抗网, 深度学习, 人脸识别, 光照归一化, 人脸光照处理

Abstract: Aiming at the influence of illumination in face recognition process, a method of illumination normalization based on CycleGAN is presented. By using the Generative Adversarial Nets and the principle of image translation, the illumination style of the darker image was shifted to the brighter image, while the surface and the structure of the face kept smooth at the same time. Unpaired data sets without labels are used, in order to achieve unsupervised removal of illumination, and greatly simplify the work of data preprocessing. Finally, the CroppedYale data set is used to train a deep learning face recognition model, using this CycleGAN model to process the test set, and comparing the accuracy before and after processing. Experiments show that this method has a strong ability to reduce the influence of human face illumination on recognition rate while basically not changing the face structure, and therefore is helpful to improve the accuracy of face recognition.

Key words: generative adversarial nets, deep learning, face recognition, illumination normalization, face illumination processing

中图分类号: 

  • TP391
[1] ?陈旭, 张军, 陈文伟, 等. 卷积网络深度学习算法与实例[J]. 广东工业大学学报, 2017, 34(6):20-26 CHEN X, ZHANG J, CHEN W W, et al. Convolutional neural network algorithm and case[J]. Journal of Guangdong University of Technology, 2017, 34(6):20-26
[2] LEE P H, WU S W, HUNG Y P. Illumination compensation using oriented local histogram equalization and its application to face recognition[J]. IEEE Transactions on Image processing, 2012, 21(9):4280-4289
[3] CHEN W, ER M J, WU S. Illumination compensation and normalization for robust face recognition using discrete cosine transform in logarithm domain[J]. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 2006, 36(2):458-466
[4] 白小叶. 光照变化下的人脸识别算法研究[D]. 南京:南京邮电大学通信与信息工程学院, 2016.
[5] 聂祥飞, 杨志军, 何雪. 利用离散余弦变换与梯度脸的人脸光照处理[J]. 微型机与应用, 2017, 36(11):50-53 NIE X F, YANG Z J, HE X. Face illumination processing using discrete cosine transform and gradient faces[J]. Microcomputer & it's applications, 2017, 36(11):50-53
[6] GOODFELLOW I J, POUGET-ABADIE J, Mirza M, et al. Generative adversarial nets[C]//International Conference on Neural Information Processing Systems. Massachusetts:MIT Press, 2014:2672-2680.
[7] RADFORD A, METZ L, CHINTALA S. Unsupervised representation learning with deep convolutional generative adversarial networks[J]. arXiv preprint arXiv:1511.06434, 2015.
[8] MIRZA M, OSINDERO S. Conditional generative adversarial nets[J]. arXiv preprint arXiv:1411.1784, 2014.
[9] ZHU J Y, PARK T, ISOLA P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks[J]. arXiv preprint arXiv:1703.10593, 2017.
[10] ISOLA P, ZHU J Y, ZHOU T H, et al. Image-to-image translation with conditional adversarial networks[J]. arXiv preprint arXiv:1611.07004, 2016.
[11] ULYANOV D, VEDALDI A, LEMPITSKY V. Instance normalization:The missing ingredient for fast stylization[J]. arXiv preprint arXiv:1607.08022, 2016.
[12] CLEVERT D A, UNTERTHINER T, HOCHREITER S. Fast and accurate deep network learning by exponential linear units (elus)[J]. arXiv preprint arXiv:1511.07289, 2015.
[13] GAO W, CAO B, SHAN S G, et al. The CAS-PEAL large-scale Chinese face database and baseline evaluations[J]. IEEE Transactions on Systems, Man, and Cybernetics-Part A:Systems and Humans, 2007, 38(1):149-161
[14] GEORGHIADES A S, BELHUMEUR P N, KRIEGMAN D J. From few to many:illumination cone models for face recognition under variable lighting and pose[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2001, 23(6):643-660
[15] LEE K C, HO J, KRIEGMAN D J. Acquiring linear subspaces for face recognition under variable lighting[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2005, 27(5):684-98
[16] WU X, HE R, SUN Z N, et al. A light CNN for deep face representation with noisy labels[J]. arXiv preprint arXiv:1511.02683, 2015.
[1] 吴俊贤, 何元烈. 基于通道注意力的自监督深度估计方法[J]. 广东工业大学学报, 2023, 40(02): 22-29.
[2] 刘冬宁, 王子奇, 曾艳姣, 文福燕, 王洋. 基于复合编码特征LSTM的基因甲基化位点预测方法[J]. 广东工业大学学报, 2023, 40(01): 1-9.
[3] 徐伟锋, 蔡述庭, 熊晓明. 基于深度特征的单目视觉惯导里程计[J]. 广东工业大学学报, 2023, 40(01): 56-60,76.
[4] 刘洪伟, 林伟振, 温展明, 陈燕君, 易闽琦. 基于MABM的消费者情感倾向识别模型——以电影评论为例[J]. 广东工业大学学报, 2022, 39(06): 1-9.
[5] 章云, 王晓东. 基于受限样本的深度学习综述与思考[J]. 广东工业大学学报, 2022, 39(05): 1-8.
[6] 郑佳碧, 杨振国, 刘文印. 基于细粒度混杂平衡的营销效果评估方法[J]. 广东工业大学学报, 2022, 39(02): 55-61.
[7] Gary Yen, 栗波, 谢胜利. 地球流体动力学模型恢复的长短期记忆网络渐进优化方法[J]. 广东工业大学学报, 2021, 38(06): 1-8.
[8] 赖峻, 刘震宇, 刘圣海. 基于全局数据混洗的小样本数据预测方法[J]. 广东工业大学学报, 2021, 38(03): 17-21.
[9] 岑仕杰, 何元烈, 陈小聪. 结合注意力与无监督深度学习的单目深度估计[J]. 广东工业大学学报, 2020, 37(04): 35-41.
[10] 陈旭, 张军, 陈文伟, 李硕豪. 卷积网络深度学习算法与实例[J]. 广东工业大学学报, 2017, 34(06): 20-26.
[11] 刘震宇, 李嘉俊, 王昆. 一种基于深度自编码器的指纹匹配定位方法[J]. 广东工业大学学报, 2017, 34(05): 15-21.
[12] 蔡正, 王福龙, 徐爱辉. 一种新的图像距离在人脸识别中的应用[J]. 广东工业大学学报, 2010, 27(3): 64-67.
[13] 黄诚; 王福龙; 刘晓亮; . 指数衰减下的KFDA方法及其人脸识别[J]. 广东工业大学学报, 2009, 26(4): 62-64.
[14] 刘晓亮; 王福龙; 黄诚; 曾爱华; . 一种加权的核Fisher鉴别分析在人脸识别中的应用[J]. 广东工业大学学报, 2009, 26(4): 65-69.
[15] 张辰; 潘保昌; 郑胜林; 卢毅涛;. 一种矢量倍增算法的神经网络人脸识别方法[J]. 广东工业大学学报, 2006, 23(3): 108-112.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!