Satellite Image Generation on Satellite Imagery
41.5FIDUCN+RemoteCLIP
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| UCN+RemoteCLIPSteps=20k, Params=1.5B, Evaluation Protocol=SD Decoder Fine-tuning, Text Encoder=RemoteCLIP2026.03 | 41.5 | 11.23 | 0.1772 | 0.5696 | 0.1406 | 0.5656 | 0.544 | 0.1344 | |
| UCN+CLIPSteps=20k, Params=1.5B, Evaluation Protocol=SD Decoder Fine-tuning, Text Encoder=CLIP2026.03 | 44.4 | 10.6 | 0.16 | 0.5558 | 0.1317 | 0.5571 | 0.5624 | 0.1265 | |
| UCN+RemoteCLIPSteps=60k, Params=1.5B, Evaluation Protocol=Text Encoders, Text Encoder=RemoteCLIP2026.03 | 48.6 | 11.23 | 0.1831 | 0.5673 | 0.1424 | 0.563 | 0.5476 | 0.1341 | |
| UCN+CLIPSteps=60k, Params=1.5B, Evaluation Protocol=Text Encoders, Text Encoder=CLIP2026.03 | 50.3 | 10.98 | 0.1645 | 0.552 | 0.1357 | 0.5576 | 0.556 | 0.1299 | |
| UCN+LongCLIPSteps=20k, Params=1.8B, Evaluation Protocol=SD Decoder Fine-tuning, Text Encoder=LongCLIP2026.03 | 57.7 | 10.85 | 0.1922 | 0.5704 | 0.1449 | 0.5508 | 0.5735 | 0.1362 | |
| UCN+LongCLIPSteps=60k, Params=1.8B, Evaluation Protocol=Text Encoders, Text Encoder=LongCLIP2026.03 | 61 | 10.97 | 0.2094 | 0.5551 | 0.1511 | 0.5502 | 0.5792 | 0.1411 | |
| UCN+CLIPSteps=20k, Params=1.5B, Evaluation Protocol=Few Steps, Text Encoder=CLIP2026.03 | 72.5 | 10.58 | 0.1665 | 0.556 | 0.1235 | 0.5545 | 0.5808 | 0.112 | |
| ControlNetSteps=60k, Params=1.4B, Evaluation Protocol=Baselines2026.03 | 76.5 | 10.94 | 0.1847 | 0.5802 | 0.1087 | 0.5603 | 0.5623 | 0.1179 | |
| Original SDSteps=0, Params=1.4B, Evaluation Protocol=Baselines, Note=No finetuning2026.03 | 86.5 | 10.56 | — | — | 0.0927 | 0.5556 | 0.7162 | 0.0978 |