? 基于条件循环生成式对抗网络的人脸图像属性变换
Journal of Computer Science and Technology
Quick Search in JCST
 Advanced Search 
      Home | PrePrint | SiteMap | Contact Us | Help
 
Indexed by   SCIE, EI ...
Bimonthly    Since 1986
Journal of Computer Science and Technology 2018, Vol. 33 Issue (3) :511-521    DOI: 10.1007/s11390-018-1835-2
Special Section of CVM 2018 << Previous Articles | Next Articles >>
基于条件循环生成式对抗网络的人脸图像属性变换
Huai-Yu Li1,2, Wei-Ming Dong1,*, Member, CCF, ACM, IEEE, Bao-Gang Hu1, Senior Member, IEEE, Member, CCF
1 National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China;
2 University of Chinese Academy of Sciences, Beijing 100049, China
Facial Image Attributes Transformation via Conditional Recycle Generative Adversarial Networks
Huai-Yu Li1,2, Wei-Ming Dong1,*, Member, CCF, ACM, IEEE, Bao-Gang Hu1, Senior Member, IEEE, Member, CCF
1 National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China;
2 University of Chinese Academy of Sciences, Beijing 100049, China

摘要
参考文献
相关文章
Download: [PDF 5422KB]  
摘要 本研究针对人脸图像属性变换提出了一种新的条件循环生成对抗网络,能够在不改变身份的情况下转换高层次的语义人脸属性。在我们的方法中,我们将输入源脸部图像输入到具有目标属性条件的条件生成器以生成具有目标属性的脸部。然后我们将生成的面循环回收到具有源属性条件的相同条件生成器。最终生成的脸部图像应该与源脸部图像具有相同的身份特征和脸部属性。因此,我们引入一个循环重建损失来强制最终生成的面部图像和源面部图像是相同的。在CelebA数据集的实验评估证明了我们方法的有效性。定性结果表明,我们的方法可以学习并生成具有指定属性的高质量且保持身份特征的面部图像。
关键词生成式对抗网络   图像编辑   人脸属性转换     
Abstract: This study introduces a novel conditional recycle generative adversarial network for facial attribute transformation, which can transform high-level semantic face attributes without changing the identity. In our approach, we input a source facial image to the conditional generator with target attribute condition to generate a face with the target attribute. Then we recycle the generated face back to the same conditional generator with source attribute condition. A face should be similar with that of the source face in personal identity and facial attributes is generated. Hence, we introduce a recycle reconstruction loss to enforce the final generated facial image and the source facial image to be identical. Evaluations on the CelebA dataset demonstrate the effectiveness of our approach. Qualitative results show that our method can learn and generate high-quality identity-preserving facial images with specified attributes.
KeywordsGenerative Adversarial Networks   Image Editing   Facial Attributes Transformation     
Received 2017-12-25;
本文基金:

This work was supported by the National Natural Science Foundation of China under Grant Nos. 61672520, 61573348, 61620106003, and 61720106006, the Beijing Natural Science Foundation of China under Grant No. 4162056, the National Key Technology Research and Development Program of China under Grant No. 2015BAH53F02, and the CASIA-Tencent YouTu Jointly Research Project. The Titan X used for this research was donated by the NVIDIA Corporation.

通讯作者: Wei-Ming Dong     Email: weiming.dong@ia.ac.cn
About author: Huai-Yu Li is a Ph.D. student in Institute of Automation, Chinese Academy of Sciences, Beijing, under the supervision of Prof. Bao-Gang Hu. He earned his Bachelor's degree in electronic information engineering from Northeast University, Shenyang, in 2014. His research interests are in artificial intelligence, computer vision, and deep learning.
引用本文:   
Huai-Yu Li, Wei-Ming Dong, Bao-Gang Hu.基于条件循环生成式对抗网络的人脸图像属性变换[J]  Journal of Computer Science and Technology , 2018,V33(3): 511-521
Huai-Yu Li, Wei-Ming Dong, Bao-Gang Hu.Facial Image Attributes Transformation via Conditional Recycle Generative Adversarial Networks[J]  Journal of Computer Science and Technology, 2018,V33(3): 511-521
链接本文:  
http://jcst.ict.ac.cn:8080/jcst/CN/10.1007/s11390-018-1835-2
Copyright 2010 by Journal of Computer Science and Technology