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

Source: [Reversible GANs for Memory-efficient Image-to-Image Translation](<https://www.sota2.com/research/paper/68fa65581e209131cea09e49>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-25T21:25:53.160Z

Authors: Tycho F.A. van der Ouderaa, Daniel E. Worrall

Published: 2019-02-07T17:03:05.000Z

## 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.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Image-to-Image Translation on in-house CT-MRI T1 (test)](<https://www.sota2.com/research/sota/image-to-image-translation-on-in-house-ct-mri-t1-test.md>): Image-to-Image Translation; in-house CT-MRI T1 (test); NRMSE: 0.0577.

- [Image-to-Image Translation on BraTS FLAIR to T1 2023 (test)](<https://www.sota2.com/research/sota/image-to-image-translation-on-brats-flair-to-t1-2023-test.md>): Image-to-Image Translation; BraTS FLAIR to T1 2023 (test); NRMSE: 0.0824.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/1902.02729>)

[PDF](<https://arxiv.org/pdf/1902.02729.pdf>)
