Classification on EuroSat (Accuracy Opt. and SAR)
81.18Accuracy (Optical)CoDe-MAE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoDe-MAEBackbone=HiViT-B, Pre-training Dataset=OSPretrain-1M, HR=true, Evaluation Protocol=10-shot linear probing2026.04 | 81.18 | 59.88 | |
| MaRSBackbone=Swin V2-B, Pre-training Dataset=MaRS-16M (32M), HR=true, Evaluation Protocol=10-shot linear probing2026.04 | 79.79 | 32.66 | |
| DOFABackbone=DOFA-B, Pre-training Dataset=DOFA-MM (8M), HR=true, Evaluation Protocol=10-shot linear probing2026.04 | 70.25 | 55.87 | |
| DeCURBackbone=ViT-S, Pre-training Dataset=SSL4EO (502K), HR=false, Evaluation Protocol=10-shot linear probing2026.04 | 60.48 | 48.08 | |
| CROMABackbone=ViT-B, Pre-training Dataset=SSL4EO (2M), HR=false, Evaluation Protocol=10-shot linear probing2026.04 | 59.12 | 47.97 | |
| SatViTBackbone=ViT-B, Pre-training Dataset=Sentinel-1/2 (2.6M), HR=false, Evaluation Protocol=10-shot linear probing2026.04 | 54.94 | 47.03 | |
| DINO-MMBackbone=ViT-S, Pre-training Dataset=BEN-MM (620K), HR=false, Evaluation Protocol=10-shot linear probing2026.04 | 50.66 | 47.39 | |
| FG-MAEBackbone=ViT-B, Pre-training Dataset=SSL4EO (2M), HR=false, Evaluation Protocol=10-shot linear probing2026.04 | 48.88 | 42.72 | |
| SwinSSLBackbone=Swin-T, Pre-training Dataset=SEN12MS (360K), HR=false, Evaluation Protocol=10-shot linear probing2026.04 | 48 | 40.85 |