# Watermark Recovery on CelebA-HQ

Source: [Watermark Recovery on CelebA-HQ](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq>)

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

Data updated: 2026-04-02T09:08:44.391Z

Evaluation of watermark recovery performance on the CelebA-HQ dataset, measured using Bit Error Rate (BER) across a comprehensive set of common and malicious distortions (e.g., noise, compression, GAN attacks).

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 | Watermark Recovery |
| Dataset (including any stated split) | CelebA-HQ |
| Metric | JpegTest BER |
| Unit | PERCENT |
| Direction | Lower is better |
| Results loaded | 8 of 8 |
| 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

- [JpegTest BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=JpegTest%20BER>): Lower is better; unit: PERCENT.

- [Resize BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=Resize%20BER>): Lower is better; unit: PERCENT.

- [GaussianBlur BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=GaussianBlur%20BER>): Lower is better; unit: PERCENT.

- [MedianBlur BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=MedianBlur%20BER>): Lower is better; unit: PERCENT.

- [Dropout BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=Dropout%20BER>): Lower is better; unit: PERCENT.

- [SaltPepper BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=SaltPepper%20BER>): Lower is better; unit: PERCENT.

- [GaussianNoise BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=GaussianNoise%20BER>): Lower is better; unit: PERCENT.

- [Avg Common Distortions BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=Avg%20Common%20Distortions%20BER>): Lower is better; unit: PERCENT.

- [FSRT BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=FSRT%20BER>): Lower is better; unit: PERCENT.

- [SimSwap BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=SimSwap%20BER>): Lower is better; unit: PERCENT.

- [CSCS BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=CSCS%20BER>): Lower is better; unit: PERCENT.

- [HifiFace BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=HifiFace%20BER>): Lower is better; unit: PERCENT.

- [Uniface BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=Uniface%20BER>): Lower is better; unit: PERCENT.

- [GANimation BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=GANimation%20BER>): Lower is better; unit: PERCENT.

- [StarGAN (Male) BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=StarGAN%20(Male)%20BER>): Lower is better; unit: PERCENT.

- [StarGAN (Young) BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=StarGAN%20(Young)%20BER>): Lower is better; unit: PERCENT.

- [StarGAN (BlackHair) BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=StarGAN%20(BlackHair)%20BER>): Lower is better; unit: PERCENT.

- [StarGAN (BlondHair) BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=StarGAN%20(BlondHair)%20BER>): Lower is better; unit: PERCENT.

- [StarGAN (BrownHair) BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=StarGAN%20(BrownHair)%20BER>): Lower is better; unit: PERCENT.

- [Avg Malicious Distortions BER](<https://www.sota2.com/research/sota/watermark-recovery-on-celeba-hq.md?metric=Avg%20Malicious%20Distortions%20BER>): Lower is better; unit: PERCENT.

## Recorded results

| Method | JpegTest BER | Result date | Report paper |
| --- | --- | --- | --- |
| LampMark | 0 | Not recorded | 68fa78c71e209131cea7a731 |
| WaveGuard | 0 | Not recorded | 68fa78c71e209131cea7a731 |
| WaveGuard | 0 | Not recorded | 68fa78c71e209131cea7a731 |
| SepMark | 0.0071 | Not recorded | 68fa78c71e209131cea7a731 |
| SepMark | 0.0071 | Not recorded | 68fa78c71e209131cea7a731 |
| FaceSigns | 0.6298 | Not recorded | 68fa78c71e209131cea7a731 |
| EditGuard | 3.5801 | Not recorded | 68fa78c71e209131cea7a731 |
| ARWGAN | 6.0034 | Not recorded | 68fa78c71e209131cea7a731 |

### LampMark

JpegTest BER: 0. Result ID: 69ce321947f34ce924b78cfb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a731.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75e31e209131cea6a7bd.md>)

Recorded evaluation configuration: {"Resolution":"256x256"}. This record does not assert independent reproduction.

### WaveGuard

JpegTest BER: 0. Result ID: 69ce321947f34ce924b78cfe.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a731.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a731.md>)

Recorded evaluation configuration: {"Resolution":"256x256","Evaluation Protocol":"Tracer"}. This record does not assert independent reproduction.

### WaveGuard

JpegTest BER: 0. Result ID: 69ce321947f34ce924b78cff.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a731.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a731.md>)

Recorded evaluation configuration: {"Resolution":"256x256","Evaluation Protocol":"Detector"}. This record does not assert independent reproduction.

### SepMark

JpegTest BER: 0.0071. Result ID: 69ce321947f34ce924b78cfc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a731.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa703c1e209131cea49cca.md>)

Recorded evaluation configuration: {"Resolution":"256x256","Evaluation Protocol":"Tracer"}. This record does not assert independent reproduction.

### SepMark

JpegTest BER: 0.0071. Result ID: 69ce321947f34ce924b78cfd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a731.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa703c1e209131cea49cca.md>)

Recorded evaluation configuration: {"Resolution":"256x256","Evaluation Protocol":"Detector"}. This record does not assert independent reproduction.

### FaceSigns

JpegTest BER: 0.6298. Result ID: 69ce321947f34ce924b78cf9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a731.md>)

Recorded evaluation configuration: {"Resolution":"256x256"}. This record does not assert independent reproduction.

### EditGuard

JpegTest BER: 3.5801. Result ID: 69ce321947f34ce924b78cfa.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a731.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71721e209131cea514c8.md>)

Recorded evaluation configuration: {"Resolution":"256x256"}. This record does not assert independent reproduction.

### ARWGAN

JpegTest BER: 6.0034. Result ID: 69ce321947f34ce924b78cf8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a731.md>)

Recorded evaluation configuration: {"Resolution":"256x256"}. This record does not assert independent reproduction.

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