# GANHopper: Multi-Hop GAN for Unsupervised Image-to-Image Translation

arXiv: https://arxiv.org/abs/2002.10102 · HTML: https://mlsanity.com/paper/2002.10102/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 24 Feb 2020.

## Built on by (most cited first, 10 of 10)

- [A survey of synthetic data augmentation methods in computer vision](https://mlsanity.com/paper/2403.10075/) (2024)
- [Sem2NeRF: Converting Single-View Semantic Masks to Neural Radiance Fields](https://mlsanity.com/paper/2203.10821/) (2022)
- [House-GAN++: Generative Adversarial Layout Refinement Networks](https://mlsanity.com/paper/2103.02574/) (2021)
- [Slice3D: Multi-Slice, Occlusion-Revealing, Single View 3D Reconstruction](https://mlsanity.com/paper/2312.02221/) (2023)
- [Physics-informed Guided Disentanglement in Generative Networks](https://mlsanity.com/paper/2107.14229/) (2021)
- [RaidaR: A Rich Annotated Image Dataset of Rainy Street Scenes](https://mlsanity.com/paper/2104.04606/) (2021)
- [UNIST: Unpaired Neural Implicit Shape Translation Network](https://mlsanity.com/paper/2112.05381/) (2021)
- [Contrastive Monotonic Pixel-Level Modulation](https://mlsanity.com/paper/2207.11517/) (2022)
- [Generative Quanta Color Imaging](https://mlsanity.com/paper/2403.19066/) (2024)
- [Image-to-Image Translation Framework Embedded with Rotation Symmetry Priors](https://mlsanity.com/paper/2604.12805/) (2026)

## Abstract (arXiv)

We introduce GANHopper, an unsupervised image-to-image translation network that transforms images gradually between two domains, through multiple hops. Instead of executing translation directly, we steer the translation by requiring the network to produce in-between images that resemble weighted hybrids between images from the input domains. Our network is trained on unpaired images from the two domains only, without any in-between images. All hops are produced using a single generator along each direction. In addition to the standard cycle-consistency and adversarial losses, we introduce a new hybrid discriminator, which is trained to classify the intermediate images produced by the generator as weighted hybrids, with weights based on a predetermined hop count. We also add a smoothness term to constrain the magnitude of each hop, further regularizing the translation. Compared to previous methods, GANHopper excels at image translations involving domain-specific image features and geometric variations while also preserving non-domain-specific features such as general color schemes.