AI-generated image detection on Midjourney generator (test)
91.35AccuracySAFE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SAFETraining Dataset=ProGAN2026.01 | 91.35 | 97.32 | |
| NPRTraining Dataset=ProGAN2026.01 | 82.4 | 91.75 | |
| Ours (ViT-L/14)Training Dataset=ProGAN, Backbone=ViT-L/142026.01 | 77.45 | 82.67 | |
| FatFormerTraining Dataset=ProGAN2026.01 | 70.7 | 82.47 | |
| LGradTraining Dataset=ProGAN2026.01 | 68.65 | 73.6 | |
| GramNetTraining Dataset=ProGAN2026.01 | 60.8 | 58.65 | |
| Ours (ViT-B/32)Training Dataset=ProGAN, Backbone=ViT-B/322026.01 | 59.35 | 74.71 | |
| UnivFDTraining Dataset=ProGAN2026.01 | 50.45 | 56.75 | |
| CNNSpotTraining Dataset=ProGAN2026.01 | 50.4 | 53.4 | |
| LNPTraining Dataset=ProGAN2026.01 | 49.95 | 49.88 | |
| FreDectTraining Dataset=ProGAN2026.01 | 43.55 | 42.94 |