Reversible GANs for Memory-efficient Image-to-Image Translation

2019/02/07 by Tycho F. A. van der Ouderaa, van der Ouderaa, Tycho F. A., Daniel E. Worrall +1 · 1 citation
Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #cs.CV

paper · pdf · doi:10.48550/arxiv.1902.02729

arxiv created 2019/02/07 · arxiv updated 2019/02/08

Abstract

The Pix2pix and CycleGAN losses have vastly improved the qualitative and quantitative visual quality of results in image-to-image translation tasks. We extend this framework by exploring approximately invertible architectures which are well suited to these losses. These architectures are approximately invertible by design and thus partially satisfy cycle-consistency before training even begins. Furthermore, since invertible architectures have constant memory complexity in depth, these models can be built arbitrarily deep. We are able to demonstrate superior quantitative output on the Cityscapes and Maps datasets at near constant memory budget.

Cited by

• DINVMark: A Deep Invertible Network for Video Watermarking
• Structure-Accurate Medical Image Translation via Dynamic Frequency Balance and Knowledge Guidance

Related

• UVCGAN: UNet Vision Transformer cycle-consistent GAN for unpaired image-to-image translation
• An Optimized Architecture for Unpaired Image-to-Image Translation
• Breaking the cycle -- Colleagues are all you need
• Chest CT Super-resolution and Domain-adaptation using Memory-efficient 3D Reversible GANs
• TraVeLGAN: Image-to-image Translation by Transformation Vector Learning