# Multi-style Image-to-Image Translation on CelebA-HQ (test)

Source: [Multi-style Image-to-Image Translation on CelebA-HQ (test)](<https://www.sota2.com/research/sota/multi-style-image-to-image-translation-on-celeba-hq-test>)

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

Data updated: 2026-02-26T06:01:09.678Z

Evaluation of Multi-style Image-to-Image Translation models on the test split of CelebA-HQ, focusing on metrics covering Realism (FID), Diversity (User Study), and Disentanglement across latent-guided, reference-guided, and gap analyses.

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.

| Scope | Recorded value |
| --- | --- |
| Task | Multi-style Image-to-Image Translation |
| Dataset (including any stated split) | CelebA-HQ (test) |
| Metric | Realism (FID) \[Reference-guided\] |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 4 of 4 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Realism (FID) \[Reference-guided\]](<https://www.sota2.com/research/sota/multi-style-image-to-image-translation-on-celeba-hq-test.md?metric=Realism%20(FID)%20%5BReference-guided%5D>): Lower is better; unit: SCALAR.

- [Realism (FID) \[Latent-guided\]](<https://www.sota2.com/research/sota/multi-style-image-to-image-translation-on-celeba-hq-test.md?metric=Realism%20(FID)%20%5BLatent-guided%5D>): Lower is better; unit: SCALAR.

- [Realism (FID) \[Gap\]](<https://www.sota2.com/research/sota/multi-style-image-to-image-translation-on-celeba-hq-test.md?metric=Realism%20(FID)%20%5BGap%5D>): Lower is better; unit: SCALAR.

- [Diversity (User Study %) \[Latent-guided\]](<https://www.sota2.com/research/sota/multi-style-image-to-image-translation-on-celeba-hq-test.md?metric=Diversity%20(User%20Study%20%25)%20%5BLatent-guided%5D>): Higher is better; unit: PERCENT.

- [Diversity (User Study %) \[Reference-guided\]](<https://www.sota2.com/research/sota/multi-style-image-to-image-translation-on-celeba-hq-test.md?metric=Diversity%20(User%20Study%20%25)%20%5BReference-guided%5D>): Higher is better; unit: PERCENT.

- [Diversity (User Study %) \[Gap\]](<https://www.sota2.com/research/sota/multi-style-image-to-image-translation-on-celeba-hq-test.md?metric=Diversity%20(User%20Study%20%25)%20%5BGap%5D>): Higher is better; unit: PERCENT.

- [Disentanglement (FID) \[Latent-guided\]](<https://www.sota2.com/research/sota/multi-style-image-to-image-translation-on-celeba-hq-test.md?metric=Disentanglement%20(FID)%20%5BLatent-guided%5D>): Lower is better; unit: SCALAR.

- [Disentanglement (FID) \[Reference-guided\]](<https://www.sota2.com/research/sota/multi-style-image-to-image-translation-on-celeba-hq-test.md?metric=Disentanglement%20(FID)%20%5BReference-guided%5D>): Lower is better; unit: SCALAR.

- [Disentanglement (FID) \[Gap\]](<https://www.sota2.com/research/sota/multi-style-image-to-image-translation-on-celeba-hq-test.md?metric=Disentanglement%20(FID)%20%5BGap%5D>): Lower is better; unit: SCALAR.

## Recorded results

| Method | Realism (FID) \[Reference-guided\] | Result date | Report paper |
| --- | --- | --- | --- |
| HiSD | 21.49 | Not recorded | 68fa69741e209131cea222ee |
| ELEGANT | 22.96 | Not recorded | 68fa69741e209131cea222ee |
| StarGANv2 | 25.49 | Not recorded | 68fa69741e209131cea222ee |
| SDIT | 33.12 | Not recorded | 68fa69741e209131cea222ee |

### HiSD

Realism (FID) \[Reference-guided\]: 21.49. Result ID: 698dab0e8b517eb43322fec2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa69741e209131cea222ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69741e209131cea222ee.md>)

[Implementation](<https://github.com/imlixinyang/HiSD>)

Recorded evaluation configuration: {"style codes":"5","style representation":"pose-invariant global style codes"}. This record does not assert independent reproduction.

### ELEGANT

Realism (FID) \[Reference-guided\]: 22.96. Result ID: 698dab0e8b517eb43322fec1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa69741e209131cea222ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa649b1e209131cea057bc.md>)

Recorded evaluation configuration: {"style codes":"5"}. This record does not assert independent reproduction.

### StarGANv2

Realism (FID) \[Reference-guided\]: 25.49. Result ID: 698dab0e8b517eb43322fec0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa69741e209131cea222ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64d91e209131cea06ae8.md>)

Recorded evaluation configuration: {"style codes":"5"}. This record does not assert independent reproduction.

### SDIT

Realism (FID) \[Reference-guided\]: 33.12. Result ID: 698dab0e8b517eb43322febf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa69741e209131cea222ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66361e209131cea0ecef.md>)

Recorded evaluation configuration: {"style codes":"5"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.