Semantic Segmentation on WHUOHS w/o domain gap
59.7mIoUSpectralX∗ScaleMAE
Evaluation results
| Method | Links | |||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SpectralX∗ScaleMAEBackbone=ViT-L, Trainable params=11.8M, Fine-tuning=PEFT2025.08 | 59.7 | 71.5 | 71.3 | 63.7 | 37.9 | 17.9 | 19.7 | 56.3 | 36.3 | 45.8 | 70.1 | 94.1 | 62.1 | 73.5 | 57.9 | 78.1 | 47.2 | 40.3 | 79.4 | 75.3 | 63 | 54.1 | 39.8 | 78.9 | 97.8 | |
| SpectralX∗SatMAE++Backbone=ViT-L, Trainable params=11.8M, Fine-tuning=PEFT2025.08 | 59.6 | 70.9 | 70.6 | 64.7 | 36.6 | 18.7 | 20.6 | 54.2 | 33.5 | 44.8 | 71.6 | 93.9 | 58.8 | 76.3 | 58.5 | 75.4 | 46.2 | 39.1 | 76.6 | 76.9 | 63.2 | 49.8 | 55.3 | 75.9 | 98.1 | |
| HyperSIGMABackbone=ViT-L, Trainable params=382.5M, Fine-tuning=Full2025.08 | 57.7 | 71.2 | 71.4 | 65 | 40.4 | 17.6 | 19.1 | 56.9 | 35.5 | 46.6 | 66.6 | 92 | 58.5 | 59.5 | 58 | 75 | 48.5 | 37.4 | 79.1 | 74.1 | 39.6 | 51.1 | 45.5 | 77.5 | 97.5 | |
| HyperSIGMAVenue=arXiv242025.08 | 57.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SatMAE++Backbone=ViT-L, Trainable params=305.5M, Fine-tuning=Full2025.08 | 55.7 | 72 | 72.5 | 66.6 | 42.4 | 19.5 | 20.4 | 54.6 | 34.9 | 39.5 | 69.4 | 87.6 | 56.1 | 38.2 | 58 | 75.8 | 51 | 38.5 | 79 | 65.6 | 40 | 45.9 | 39.3 | 73 | 95.9 | |
| SatMAE++Venue=CVPR242025.08 | 55.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSTCBackbone=PVTv2-B1, Trainable params=25.7M2025.08 | 54.5 | 71.5 | 71.3 | 63.7 | 37.9 | 17.9 | 19.7 | 56.3 | 36.2 | 45.8 | 70.5 | 93.3 | 61.4 | 73.3 | 57.9 | 78 | 47.2 | 40.1 | 78.9 | 75.3 | 63 | 53.8 | 39.8 | 78.7 | 97.1 | |
| DAENBackbone=RefineNet-lw, Trainable params=48.7M2025.08 | 54 | 70.1 | 70.8 | 64.8 | 40.3 | 13.3 | 4.3 | 48.7 | 32.3 | 40.5 | 62 | 85.7 | 52.5 | 48.7 | 54.8 | 76.6 | 53.6 | 34.5 | 76.5 | 69.9 | 31.4 | 48.3 | 42.9 | 76.9 | 97.2 | |
| ScaleMAEVenue=ICCV232025.08 | 53.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ScaleMAEBackbone=ViT-L, Trainable params=303.4M, Fine-tuning=Full2025.08 | 53.4 | 66.8 | 65.8 | 61.6 | 38.8 | 16.9 | 18.4 | 52.8 | 32.8 | 38.7 | 67.3 | 85.4 | 54 | 49.9 | 56.7 | 71.3 | 42.3 | 32 | 77 | 67.2 | 36.3 | 41.3 | 45.8 | 72.2 | 90.7 | |
| SLR∗ScaleMAEBackbone=ViT-L, Trainable params=7.3M, Fine-tuning=PEFT2025.08 | 52.4 | 69.4 | 69.2 | 62.1 | 32.4 | 17.7 | 16.4 | 54.8 | 30.6 | 40.9 | 66.6 | 77.1 | 56.9 | 45.3 | 58 | 75.9 | 44.8 | 37.8 | 73.6 | 59.3 | 34.2 | 50.3 | 35.6 | 62.9 | 85.3 | |
| LoRA∗ScaleMAEBackbone=ViT-L, Trainable params=6.9M, Fine-tuning=PEFT2025.08 | 52.1 | 68.5 | 69 | 62.8 | 35.4 | 17.4 | 17.6 | 54 | 31.4 | 40.6 | 66.4 | 75.4 | 52.9 | 51.9 | 54.9 | 76.7 | 45.7 | 37.7 | 76.6 | 57.7 | 24.6 | 44.2 | 39.7 | 63.8 | 85.6 | |
| SpectralGPT+Venue=TPAMI242025.08 | 49.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpectralGPT+Backbone=ViT-L, Trainable params=307.2M, Fine-tuning=Full2025.08 | 49.3 | 63.4 | 63 | 58.7 | 32.6 | 17.2 | 14.4 | 50.1 | 29.1 | 39.1 | 52.8 | 84.5 | 44.8 | 39.2 | 49 | 69.4 | 37.7 | 31.5 | 70.4 | 65.7 | 54.7 | 35.1 | 36.2 | 56.8 | 88.6 | |
| LoRA∗SatMAE++Backbone=ViT-L, Trainable params=6.9M, Fine-tuning=PEFT2025.08 | 48.6 | 67 | 66.2 | 60.3 | 35.9 | 16.7 | 11.4 | 49.8 | 29 | 38.9 | 62.2 | 74.6 | 55.2 | 24.9 | 58.3 | 73.8 | 41.3 | 30.7 | 76.4 | 52.8 | 28.2 | 41.6 | 23.7 | 67.1 | 79 | |
| SLR∗SatMAE++Backbone=ViT-L, Trainable params=7.3M, Fine-tuning=PEFT2025.08 | 48 | 66.6 | 66.2 | 59.8 | 38.1 | 13.5 | 13.4 | 54.1 | 29.8 | 39.2 | 63 | 70.5 | 48.1 | 32.9 | 55.6 | 73.6 | 41.4 | 35.4 | 74.3 | 51.3 | 13.5 | 48.3 | 23.3 | 62.8 | 77.1 | |
| SSCBackbone=ResNet101, Trainable params=50.4M2025.08 | 45.7 | 70.4 | 69.1 | 62.5 | 38.6 | 0.63 | 0 | 47.8 | 32.1 | 37.2 | 61.3 | 89.5 | 47.9 | 0 | 47.3 | 75.1 | 47.6 | 24.5 | 69.4 | 71.1 | 9.6 | 26.1 | 0 | 72.7 | 95.9 |