# Mind the Gap: Domain Gap Control for Single Shot Domain Adaptation for Generative Adversarial Networks

Source: [Mind the Gap: Domain Gap Control for Single Shot Domain Adaptation for Generative Adversarial Networks](<https://www.sota2.com/research/paper/68fa6b521e209131cea2da09>)

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

Data updated: 2025-10-25T23:46:07.580Z

Authors: Peihao Zhu, Rameen Abdal, John Femiani, Peter Wonka

Published: 2021-10-15T22:32:12.000Z

## Abstract

We present a new method for one shot domain adaptation. The input to our method is trained GAN that can produce images in domain A and a single reference image I\_B from domain B. The proposed algorithm can translate any output of the trained GAN from domain A to domain B. There are two main advantages of our method compared to the current state of the art: First, our solution achieves higher visual quality, e.g. by noticeably reducing overfitting. Second, our solution allows for more degrees of freedom to control the domain gap, i.e. what aspects of image I\_B are used to define the domain B. Technically, we realize the new method by building on a pre-trained StyleGAN generator as GAN and a pre-trained CLIP model for representing the domain gap. We propose several new regularizers for controlling the domain gap to optimize the weights of the pre-trained StyleGAN generator to output images in domain B instead of domain A. The regularizers prevent the optimization from taking on too many attributes of the single reference image. Our results show significant visual improvements over the state of the art as well as multiple applications that highlight improved control.

## 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.

- [Facial Stylization on Style Reference Example-1](<https://www.sota2.com/research/sota/facial-stylization-on-style-reference-example-1.md>): Facial Stylization; Style Reference Example-1; LPIPS: 0.34.

- [Facial Stylization on Style Reference Example-3](<https://www.sota2.com/research/sota/facial-stylization-on-style-reference-example-3.md>): Facial Stylization; Style Reference Example-3; LPIPS: 0.328.

- [Facial Stylization on Style Reference Example-2](<https://www.sota2.com/research/sota/facial-stylization-on-style-reference-example-2.md>): Facial Stylization; Style Reference Example-2; LPIPS: 0.361.

- [Domain Adaptation on AFHQ Cat to Fox](<https://www.sota2.com/research/sota/domain-adaptation-on-afhq-cat-to-fox.md>): Domain Adaptation; AFHQ Cat to Fox; FID: 82.3.

- [Image Stylization on OSASIS User Study](<https://www.sota2.com/research/sota/image-stylization-on-osasis-user-study.md>): Image Stylization; OSASIS User Study; Selection Ratio: 10.

- [Domain Adaptation on FFHQ One-shot Domain Adaptation 1.0 (val)](<https://www.sota2.com/research/sota/domain-adaptation-on-ffhq-one-shot-domain-adaptation-1-0-val.md>): Domain Adaptation; FFHQ One-shot Domain Adaptation 1.0 (val); KID (Amedeo Modigliani): 146.84.

- [Domain Adaptation on AFHQ Cat to Tiger](<https://www.sota2.com/research/sota/domain-adaptation-on-afhq-cat-to-tiger.md>): Domain Adaptation; AFHQ Cat to Tiger; FID: 46.72.

- [Domain Adaptation on AFHQ Cat to Wolf](<https://www.sota2.com/research/sota/domain-adaptation-on-afhq-cat-to-wolf.md>): Domain Adaptation; AFHQ Cat to Wolf; FID: 58.65.

- [Domain Adaptation on Cat Domain Adaptation One-shot 1.0 (val)](<https://www.sota2.com/research/sota/domain-adaptation-on-cat-domain-adaptation-one-shot-1-0-val.md>): Domain Adaptation; Cat Domain Adaptation One-shot 1.0 (val); KID (Tiger): 48.27.

- [Domain Adaptation on FFHQ to Amedeo Modigliani](<https://www.sota2.com/research/sota/domain-adaptation-on-ffhq-to-amedeo-modigliani.md>): Domain Adaptation; FFHQ to Amedeo Modigliani; FID: 215.88.

- [Domain Adaptation on FFHQ to Fernand Leger](<https://www.sota2.com/research/sota/domain-adaptation-on-ffhq-to-fernand-leger.md>): Domain Adaptation; FFHQ to Fernand Leger; FID: 278.46.

- [Domain Adaptation on FFHQ to Raphael](<https://www.sota2.com/research/sota/domain-adaptation-on-ffhq-to-raphael.md>): Domain Adaptation; FFHQ to Raphael; FID: 193.76.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2110.08398>)

[PDF](<https://arxiv.org/pdf/2110.08398.pdf>)