Related papers: Augmented Cyclic Consistency Regularization for Unpaired Image-to-Image Translation

Augmented Cyclic Consistency Regularization for Unpaired Image-to-Image Translation

URL: http://arxiv.org/abs/2003.00187v2
Date: Mon, 12 Oct 2020 16:07:23 GMT
Title: Augmented Cyclic Consistency Regularization for Unpaired Image-to-Image Translation
Authors: Takehiko Ohkawa, Naoto Inoue, Hirokatsu Kataoka, Nakamasa Inoue
Abstract summary: Augmented Cyclic Consistency Regularization (A CCR) is a novel regularization method for unpaired I2I translation. Our method outperforms the consistency regularized GAN (CR-GAN) in real-world translations.
Score: 22.51574923085135
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Abstract: Unpaired image-to-image (I2I) translation has received considerable attention in pattern recognition and computer vision because of recent advancements in generative adversarial networks (GANs). However, due to the lack of explicit supervision, unpaired I2I models often fail to generate realistic images, especially in challenging datasets with different backgrounds and poses. Hence, stabilization is indispensable for GANs and applications of I2I translation. Herein, we propose Augmented Cyclic Consistency Regularization (ACCR), a novel regularization method for unpaired I2I translation. Our main idea is to enforce consistency regularization originating from semi-supervised learning on the discriminators leveraging real, fake, reconstructed, and augmented samples. We regularize the discriminators to output similar predictions when fed pairs of original and perturbed images. We qualitatively clarify why consistency regularization on fake and reconstructed samples works well. Quantitatively, our method outperforms the consistency regularized GAN (CR-GAN) in real-world translations and demonstrates efficacy against several data augmentation variants and cycle-consistent constraints.