Scene Classification on EuroSAT-MS
96.2Overall AccuracyGFM-Swin
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GFM-SwinEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 96.2 | 96.2 | |
| RemoteCLIPEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 95.1 | 95.1 | |
| TerraMind-LEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 93.2 | 93.2 | |
| FLOROEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 91 | 90.9 | |
| DOFAEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 90.4 | 90.4 | |
| CROMAEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 90.3 | 90.3 | |
| SatlasNetEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 90.3 | 90.2 | |
| SpectralGPTEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 89 | 89 | |
| Scale-MAEEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 85.8 | 85.8 | |
| ViT BaselineTraining Protocol=Trained from scratch, Training Epochs=50, Batch Size=82026.05 | 81 | 80.5 | |
| SSL4EO-MoCoEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 80.8 | 80.4 | |
| Prithvi 1.0 100MEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 78.9 | 78.6 | |
| SSL4EO-Data2VecEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 78.8 | 78.5 | |
| SSL4EO-MAEEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 77.6 | 77.4 | |
| SSL4EO-DINOEvaluation Protocol=PANGAEA frozen-encoder, Downstream Adaptation=Lightweight linear classification head, Training Epochs=50, Batch Size=82026.05 | 76.2 | 75.9 |