Facial Animation Using CycleGAN
Quan Li, Haiyi Zhang · 2021 International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME) · 2021
This paper mainly shows the image translation from people's faces to anime characters by using Cycle-Consistent Adversarial Network (CycleGAN). Generative adversarial networks (GANs) have become a major field of deep learning in recent years. CycleGAN, as a variant of GAN, has made profound achievements in image translation problems. Experiments in this paper demonstrate that CycleGAN can generate realistic results of the transformation between real faces and anime faces under unsupervised learning.