Image Classification on fMoW (val)
79.3AccuracyNeighborMAE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| NeighborMAEPre-training Data=fMoW [8], Evaluation Protocol=full fine-tuning2026.03 | 79.3 | — | |
| ScaleMAEPre-training Data=fMoW [8], Evaluation Protocol=full fine-tuning2026.03 | 78.8 | — | |
| MAEPre-training Data=fMoW [8], Evaluation Protocol=full fine-tuning2026.03 | 78.2 | — | |
| DOFAPre-training Data=DOFA [46], Evaluation Protocol=full fine-tuning2026.03 | 78 | — | |
| NeighborMAEPre-training Data=Satl. [41], Evaluation Protocol=full fine-tuning2026.03 | 77.9 | — | |
| Temporal SatMAEPre-training Data=fMoW [8], Evaluation Protocol=full fine-tuning2026.03 | 77.6 | — | |
| SatMAE++Pre-training Data=fMoW [8], Evaluation Protocol=full fine-tuning2026.03 | 77.4 | — | |
| SatMAEPre-training Data=fMoW [8], Evaluation Protocol=full fine-tuning2026.03 | 76.9 | — | |
| MAEPre-training Data=Satl. [41], Evaluation Protocol=full fine-tuning2026.03 | 76.7 | — | |
| MoCo-V2+TPBackbone=ResNet50, Evaluation Protocol=Finetune2020.11 | 71.55 | 67.34 | |
| CrossScale MAEPre-training Data=fMoW [8], Evaluation Protocol=full fine-tuning2026.03 | 71.4 | — | |
| MoCo-V2+Geo+TPBackbone=ResNet50, Evaluation Protocol=Finetune2020.11 | 70.6 | 66.56 | |
| Sup. Learning (IN wts. init.)Backbone=ResNet50, Evaluation Protocol=Finetune2020.11 | 69.07 | 64.72 | |
| Sup. Learning (Scratch)Backbone=ResNet50, Evaluation Protocol=Finetune2020.11 | 69.05 | 64.71 | |
| MoCo-V2+GeoBackbone=ResNet50, Evaluation Protocol=Finetune2020.11 | 69.04 | 66.6 | |
| NeighborMAEPre-training Data=fMoW [8], Evaluation Protocol=frozen backbone2026.03 | 68.8 | — | |
| MoCo-V2+TPBackbone=ResNet50, Evaluation Protocol=Frozen2020.11 | 68.32 | 64.53 | |
| ScaleMAEPre-training Data=fMoW [8], Evaluation Protocol=frozen backbone2026.03 | 67.1 | — | |
| MAEPre-training Data=fMoW [8], Evaluation Protocol=frozen backbone2026.03 | 66.8 | — | |
| SatMAE++Pre-training Data=fMoW [8], Evaluation Protocol=frozen backbone2026.03 | 66.4 | — | |
| MoCo-V2+Geo+TPBackbone=ResNet50, Evaluation Protocol=Frozen2020.11 | 66.33 | 63.13 | |
| MoCo-V2Backbone=ResNet50, Evaluation Protocol=Finetune2020.11 | 64.34 | 60.61 | |
| SatMAEPre-training Data=fMoW [8], Evaluation Protocol=frozen backbone2026.03 | 64.3 | — | |
| MoCo-V2+GeoBackbone=ResNet50, Evaluation Protocol=Frozen2020.11 | 64.07 | 61.6 | |
| MoCo-V2 (pre. on IN)Backbone=ResNet50, Evaluation Protocol=Finetune2020.11 | 62.9 | 57.36 | |
| DOFAPre-training Data=DOFA [46], Evaluation Protocol=frozen backbone2026.03 | 62.6 | — | |
| MoCo-V2Backbone=ResNet50, Evaluation Protocol=Frozen2020.11 | 60.69 | 55.47 | |
| NeighborMAEPre-training Data=Satl. [41], Evaluation Protocol=frozen backbone2026.03 | 58.8 | — | |
| Temporal SatMAEPre-training Data=fMoW [8], Evaluation Protocol=frozen backbone2026.03 | 57.6 | — | |
| MAEPre-training Data=Satl. [41], Evaluation Protocol=frozen backbone2026.03 | 57.3 | — | |
| Geoloc. LearningBackbone=ResNet50, Evaluation Protocol=Finetune2020.11 | 56.59 | 52.23 | |
| Geoloc. LearningBackbone=ResNet50, Evaluation Protocol=Frozen2020.11 | 52.4 | 48.96 | |
| CrossScale MAEPre-training Data=fMoW [8], Evaluation Protocol=frozen backbone2026.03 | 42.6 | — | |
| MoCo-V2 (pre. on IN)Backbone=ResNet50, Evaluation Protocol=Frozen2020.11 | 37.05 | 31.55 |