# Deepfake Detection on NT, DF, F2F, FS, FSH, CDFv2, and FFIW10K (test)

Source: [Deepfake Detection on NT, DF, F2F, FS, FSH, CDFv2, and FFIW10K (test)](<https://www.sota2.com/research/sota/deepfake-detection-on-nt-df-f2f-fs-fsh-cdfv2-and-ffiw10k-test>)

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

Data updated: 2026-02-26T05:53:31.930Z

Evaluation of Deepfake Detection performance across a suite of seven distinct datasets: NeuralTextures (NT), Deepfakes (DF), Face2Face (F2F), FaceSwap (FS), FaceShifter (FSH), CelebDFV2 (CDFv2), and Face Forensics in the Wild (FFIW10K) raw test sets.

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 | Deepfake Detection |
| Dataset (including any stated split) | NeuralTextures (NT), Deepfakes (DF), Face2Face (F2F), FaceSwap (FS), FaceShifter (FSH), CelebDFV2 (CDFv2), and Face Forensics in the Wild (FFIW10K) raw (test) |
| Metric | NT |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 10 of 10 |
| 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

- [NT](<https://www.sota2.com/research/sota/deepfake-detection-on-nt-df-f2f-fs-fsh-cdfv2-and-ffiw10k-test.md?metric=NT>): Higher is better; unit: PERCENT.

- [DF](<https://www.sota2.com/research/sota/deepfake-detection-on-nt-df-f2f-fs-fsh-cdfv2-and-ffiw10k-test.md?metric=DF>): Higher is better; unit: PERCENT.

- [F2F](<https://www.sota2.com/research/sota/deepfake-detection-on-nt-df-f2f-fs-fsh-cdfv2-and-ffiw10k-test.md?metric=F2F>): Higher is better; unit: PERCENT.

- [FS](<https://www.sota2.com/research/sota/deepfake-detection-on-nt-df-f2f-fs-fsh-cdfv2-and-ffiw10k-test.md?metric=FS>): Higher is better; unit: PERCENT.

- [FSH](<https://www.sota2.com/research/sota/deepfake-detection-on-nt-df-f2f-fs-fsh-cdfv2-and-ffiw10k-test.md?metric=FSH>): Higher is better; unit: PERCENT.

- [CDFv2](<https://www.sota2.com/research/sota/deepfake-detection-on-nt-df-f2f-fs-fsh-cdfv2-and-ffiw10k-test.md?metric=CDFv2>): Higher is better; unit: PERCENT.

- [FFIW10K](<https://www.sota2.com/research/sota/deepfake-detection-on-nt-df-f2f-fs-fsh-cdfv2-and-ffiw10k-test.md?metric=FFIW10K>): Higher is better; unit: PERCENT.

- [Average Performance](<https://www.sota2.com/research/sota/deepfake-detection-on-nt-df-f2f-fs-fsh-cdfv2-and-ffiw10k-test.md?metric=Average%20Performance>): Higher is better; unit: PERCENT.

## Recorded results

| Method | NT | Result date | Report paper |
| --- | --- | --- | --- |
| BZNet | 79 | Not recorded | 68fa708c1e209131cea4ba52 |
| SBIs | 77.75 | Not recorded | 68fa708c1e209131cea4ba52 |
| QAD-E | 76.27 | Not recorded | 68fa708c1e209131cea4ba52 |
| ADD | 75.84 | Not recorded | 68fa708c1e209131cea4ba52 |
| Fang & Lin | 75.49 | Not recorded | 68fa708c1e209131cea4ba52 |
| QAD-R | 75.18 | Not recorded | 68fa708c1e209131cea4ba52 |
| F3Net | 70.95 | Not recorded | 68fa708c1e209131cea4ba52 |
| MesoNet | 70.23 | Not recorded | 68fa708c1e209131cea4ba52 |
| Rössler et al. | 69.89 | Not recorded | 68fa708c1e209131cea4ba52 |
| MAT | 69.53 | Not recorded | 68fa708c1e209131cea4ba52 |

### BZNet

NT: 79. Result ID: 699133f83a6c3148490672b3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

Recorded evaluation configuration: {"Compression":"Random Image Compression (JPEG)","Setting":"Quality-agnostic","Source":"Pre-trained weights"}. This record does not assert independent reproduction.

### SBIs

NT: 77.75. Result ID: 699133f83a6c3148490672b2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c5d1e209131cea3399c.md>)

[Implementation](<https://github.com/mapooon/SelfBlendedImages>)

Recorded evaluation configuration: {"Compression":"Random Image Compression (JPEG)","Setting":"Quality-agnostic","Source":"Pre-trained weights"}. This record does not assert independent reproduction.

### QAD-E

NT: 76.27. Result ID: 699133f83a6c3148490672b6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

Recorded evaluation configuration: {"Compression":"Random Image Compression (JPEG)","Setting":"Quality-agnostic","Backbone":"EfficientNet"}. This record does not assert independent reproduction.

### ADD

NT: 75.84. Result ID: 699133f83a6c3148490672b4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b5e1e209131cea2dc27.md>)

Recorded evaluation configuration: {"Compression":"Random Image Compression (JPEG)","Setting":"Quality-agnostic","Source":"Pre-trained weights"}. This record does not assert independent reproduction.

### Fang & Lin

NT: 75.49. Result ID: 699133f83a6c3148490672b1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a051e209131cea257b5.md>)

Recorded evaluation configuration: {"Compression":"Random Image Compression (JPEG)","Setting":"Quality-agnostic"}. This record does not assert independent reproduction.

### QAD-R

NT: 75.18. Result ID: 699133f83a6c3148490672b5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

Recorded evaluation configuration: {"Compression":"Random Image Compression (JPEG)","Setting":"Quality-agnostic","Backbone":"ResNet-50"}. This record does not assert independent reproduction.

### F3Net

NT: 70.95. Result ID: 699133f83a6c3148490672af.

[Reporting paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa689c1e209131cea1d080.md>)

Recorded evaluation configuration: {"Compression":"Random Image Compression (JPEG)","Setting":"Quality-agnostic"}. This record does not assert independent reproduction.

### MesoNet

NT: 70.23. Result ID: 699133f83a6c3148490672ad.

[Reporting paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69cc1e209131cea241fa.md>)

Recorded evaluation configuration: {"Compression":"Random Image Compression (JPEG)","Setting":"Quality-agnostic"}. This record does not assert independent reproduction.

### Rössler et al.

NT: 69.89. Result ID: 699133f83a6c3148490672ae.

[Reporting paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66421e209131cea0f418.md>)

[Implementation](<https://github.com/ondyari/FaceForensics>)

Recorded evaluation configuration: {"Compression":"Random Image Compression (JPEG)","Setting":"Quality-agnostic"}. This record does not assert independent reproduction.

### MAT

NT: 69.53. Result ID: 699133f83a6c3148490672b0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa708c1e209131cea4ba52.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa696f1e209131cea22082.md>)

Recorded evaluation configuration: {"Compression":"Random Image Compression (JPEG)","Setting":"Quality-agnostic"}. This record does not assert independent reproduction.

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