Cloud detection on L8 Biome Landsat 8 OLI TIRS TOA (test)
48.77AccuracySimCLR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SimCLRBackbone=ResNet-50, Pre-training=SSL4EO-L SimCLR, Architecture=U-Net, Protocol=Frozen backbone2023.06 | 48.77 | 32.41 | |
| ImageNetBackbone=ViT-S16, Pre-training=ImageNet supervised, Architecture=U-Net, Protocol=Frozen backbone2023.06 | 47.29 | 30.98 | |
| MoCoBackbone=ViT-S16, Pre-training=SSL4EO-L MoCo, Architecture=U-Net, Protocol=Frozen backbone2023.06 | 46.66 | 30.33 | |
| MoCoBackbone=ResNet-50, Pre-training=SSL4EO-L MoCo, Architecture=U-Net, Protocol=Frozen backbone2023.06 | 45.95 | 29.23 | |
| ImageNetBackbone=ResNet-50, Pre-training=ImageNet supervised, Architecture=U-Net, Protocol=Frozen backbone2023.06 | 45.73 | 29.78 | |
| MoCoBackbone=ResNet-18, Pre-training=SSL4EO-L MoCo, Architecture=U-Net, Protocol=Frozen backbone2023.06 | 42.7 | 27.33 | |
| SimCLRBackbone=ViT-S16, Pre-training=SSL4EO-L SimCLR, Architecture=U-Net, Protocol=Frozen backbone2023.06 | 42.33 | 26.99 | |
| ImageNetBackbone=ResNet-18, Pre-training=ImageNet supervised, Architecture=U-Net, Protocol=Frozen backbone2023.06 | 41.86 | 24.67 | |
| SimCLRBackbone=ResNet-18, Pre-training=SSL4EO-L SimCLR, Architecture=U-Net, Protocol=Frozen backbone2023.06 | 39.17 | 24.44 |