AI-generated image detection on Git-Rand-15k (test)
99.95F1 Score (Fake Class)VQVAE
Evaluation results
| Method | Links | |
|---|---|---|
| VQVAECondition=T2026.03 | 99.95 | |
| GeoSynthCondition=T2026.03 | 99.81 | |
| GeoSynth-OSMCondition=T + O2026.03 | 99.68 | |
| PixArt-α-XL/2Condition=T2026.03 | 98.83 | |
| VQGANCondition=T2026.03 | 97.08 | |
| PixArt-Σ-XL/2Condition=T2026.03 | 97.03 | |
| CRS-DiffCondition=T2026.03 | 96.55 | |
| SD 3Condition=T2026.03 | 96.22 | |
| GeoRSSDCondition=T2026.03 | 95.24 | |
| DiffusionSatCondition=T + L2026.03 | 95.04 | |
| SDXLCondition=T2026.03 | 92.94 | |
| Text2EarthCondition=T2026.03 | 77.93 | |
| GeoDiT-XL/2-αCondition=T2026.03 | 74.25 | |
| GeoDiT-XL/2-ΣCondition=T + P + L2026.03 | 72.66 |