Image Classification on EuroSAT RGB (test)
98.94AccuracySatMAE
Evaluation results
| Method | Links | |
|---|---|---|
| SatMAEbackbone=ViT-L, params=~300M, pretrain=RS foundation, img=224, evaluation protocol=linear-probe/KNN-52026.05 | 98.94 | |
| SeCobackbone=ResNet-18, params=11M, pretrain=RS foundation, img=224, evaluation protocol=linear-probe/KNN-52026.05 | 96.3 | |
| LeJEPA(LSIGReg)backbone=resnet18 lowres, params=11M, pretrain=in-domain SSL, img=64→128, evaluation protocol=linear probe2026.05 | 96.11 | |
| Scale-MAEbackbone=ViT-L, params=~300M, pretrain=RS foundation, img=64, evaluation protocol=linear-probe/KNN-52026.05 | 96 | |
| UR-JEPA(LCGLT)backbone=resnet18 lowres, params=11M, pretrain=in-domain SSL, img=64→128, evaluation protocol=linear probe2026.05 | 95.97 | |
| MoCo-ImageNetbackbone=ResNet-50, params=25M, pretrain=ImageNet SSL, img=224, evaluation protocol=linear-probe/KNN-52026.05 | 95.76 | |
| MoCo-ImageNetbackbone=ResNet-50, params=25M, pretrain=ImageNet SSL, img=64, evaluation protocol=linear-probe/KNN-52026.05 | 94.11 | |
| SeCobackbone=ResNet-18, params=11M, pretrain=RS foundation, img=64, evaluation protocol=linear-probe/KNN-52026.05 | 93.14 | |
| ImageNet-supbackbone=ResNet-50, params=25M, pretrain=ImageNet, img=224, evaluation protocol=linear-probe/KNN-52026.05 | 93.13 | |
| GASSLbackbone=ResNet-18, params=11M, pretrain=RS foundation, img=64, evaluation protocol=linear-probe/KNN-52026.05 | 89.51 | |
| ImageNet-supbackbone=ResNet-50, params=25M, pretrain=ImageNet, img=64, evaluation protocol=linear-probe/KNN-52026.05 | 82.09 |