Face Attribute Editing using AttGAN and Guide Mask
Hyeon Seok Yang, Young Shik Moon · 2019 International Conference on Electronics, Information, and Communication (ICEIC) · 2019
Recently, face attribute editing researches using GAN have shown excellent performance. AttGAN, one of the most recent studies, has suggested a way to change only the desired attributes. However, unnatural distortion may occur in the background area when changing attributes such as hair color or baldhead. In this paper, we propose a method to remove the background distortion and make the result more natural by using a mask made by human parsing algorithm and a mask made by hand. Experiments with CelebA datasets show that this method can reduce distortion caused by existing methods.