Table 6.
Class similarity of MLRSNet, Resisc45 and PatternNet to the downstream tasks. PD and DT stand for pretraining dataset and downstream task, respectively.
| DT → |
AID | EuroSAT | UCM |
|---|---|---|---|
| PD↓ | |||
| MLRSNet | 66.6 % | 40 % | 76.1 % |
| PatternNet | 30 % | 30 % | 85.71 % |
| Resisc45 | 60 % | 40 % | 90.47 % |