AI-generated image detection on VQDM generator (test)
94AccuracyFatFormer
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FatFormerTraining Dataset=ProGAN2026.01 | 94 | 98.99 | |
| Ours (ViT-L/14)Training Dataset=ProGAN, Backbone=ViT-L/142026.01 | 91.95 | 99.18 | |
| SAFETraining Dataset=ProGAN2026.01 | 90.55 | 98.77 | |
| Ours (ViT-B/32)Training Dataset=ProGAN, Backbone=ViT-B/322026.01 | 88.7 | 97.59 | |
| UnivFDTraining Dataset=ProGAN2026.01 | 85.25 | 97.34 | |
| FreDectTraining Dataset=ProGAN2026.01 | 78.4 | 85.77 | |
| NPRTraining Dataset=ProGAN2026.01 | 72.05 | 78.38 | |
| LGradTraining Dataset=ProGAN2026.01 | 69.85 | 73.65 | |
| LNPTraining Dataset=ProGAN2026.01 | 64.5 | 70.15 | |
| GramNetTraining Dataset=ProGAN2026.01 | 59.15 | 55.18 | |
| CNNSpotTraining Dataset=ProGAN2026.01 | 53.05 | 55.02 |