Deepfake detection on OpenForensics (test)
99.22AccuracyDeiTFake
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| DeiTFakeMethodology=Dual-Stage Trained DeiT Model with Complex Augmentations2025.11 | 99.22 | 99.97 | — | — | — | — | — | |
| HiFEMethodology=High-frequency enhancement network for compressed content2025.11 | 99.03 | 99.9 | — | — | — | — | — | |
| DAW-FDD + FFTBackbone=MobileNetV3-Small, Category=In-Processing Methods2026.06 | 96.48 | 99.44 | 5.7 | 11.58 | 17.28 | 94.53 | 14 | |
| DAW-FDDBackbone=MobileNetV3-Small, Category=In-Processing Methods2026.06 | 96.33 | 99.38 | 4.91 | 11.95 | 16.86 | 93.9 | 14 | |
| DAG-FDD + FFTBackbone=MobileNetV3-Small, Category=In-Processing Methods2026.06 | 96.32 | 99.37 | 5.67 | 11.94 | 17.61 | 94.67 | 14 | |
| PG-FDD + FFT⋆Backbone=MobileNetV3-Small, Category=In-Processing Methods2026.06 | 96.3 | 99.33 | 4.04 | 12.02 | 16.06 | 94.62 | 14 | |
| DAG-FDDBackbone=MobileNetV3-Small, Category=In-Processing Methods2026.06 | 95.86 | 99.24 | 4.31 | 12.47 | 16.78 | 94.35 | 14 | |
| FFTBackbone=MobileNetV3-Small, Category=Post-Processing Methods2026.06 | 95.81 | 99.19 | 5.4 | 14.86 | 20.26 | 93.95 | 14 | |
| DAID + FFTBackbone=MobileNetV3-Small, Category=In-Processing Methods2026.06 | 95.12 | 99.01 | 6.57 | 15.48 | 22.05 | 92.87 | 14 | |
| Tanfoni et al. (2024)Methodology=Facial Segmentation + transfer learning (DeepLabV3+)2025.11 | 95.04 | 96 | — | — | — | — | — | |
| PG-FDDBackbone=MobileNetV3-Small, Category=In-Processing Methods2026.06 | 94.86 | 99.12 | 5.18 | 10.72 | 15.9 | 92.51 | 13 | |
| DAIDBackbone=MobileNetV3-Small, Category=In-Processing Methods2026.06 | 94.42 | 98.76 | 9.01 | 14.7 | 23.71 | 92.69 | 12 | |
| FF DiscoverBackbone=MobileNetV3-Small, Category=Post-Processing Methods2026.06 | 93.62 | — | 10.65 | 15 | 25.66 | 91.91 | 9 | |
| Fair-FLIPBackbone=MobileNetV3-Small, Category=Post-Processing Methods2026.06 | 92.95 | 98.63 | 4.96 | 15.63 | 20.59 | 91.8 | 11 | |
| BASEBackbone=MobileNetV3-Small, Category=REFERENCE2026.06 | 92.88 | 98.54 | 5.71 | 15.84 | 21.55 | 91.66 | — | |
| BPFABackbone=MobileNetV3-Small, Category=Post-Processing Methods2026.06 | 92.87 | 98.54 | 5.71 | 15.84 | 21.54 | 91.66 | 14 | |
| FILTERMethodology=Multi-face relationship learning + global feature aggregation2025.11 | 92.04 | 98 | — | — | — | — | — | |
| FF MaxBackbone=MobileNetV3-Small, Category=Post-Processing Methods2026.06 | 91.86 | — | 15.91 | 12.6 | 28.51 | 89.91 | 6 | |
| COMICSMethodology=Bi-grained contrastive learning with integrated face extraction2025.11 | 88.2 | — | — | — | — | — | — |