Frequency Domain Image Translation: More Photo-realistic, Better Identity-preserving

• Published in: Proceedings / IEEE International Conference on Computer Vision pp. 13910 - 13920
• Main Authors: Cai, Mu, Zhang, Hong, Huang, Huijuan, Geng, Qichuan, Li, Yixuan, Huang, Gao
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.01.2021
• Subjects: Computer vision; Frequency-domain analysis; Image and video synthesis; Image synthesis; Neural generative models; Task analysis; Training; Visualization

Image-to-image translation has been revolutionized with GAN-based methods. However, existing methods lack the ability to preserve the identity of the source domain. As a result, synthesized images can often over-adapt to the reference domain, losing important structural characteristics and suffering...