Building Extraction on xBD pre
77.99IoUPi-Seg
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Pi-SegBackbone=ViT-L, Type=OVRSIS2026.04 | 77.99 | 90.15 | |
| GSNETBackbone=ViT-L, Type=OVRSIS2026.04 | 76.23 | 91.72 | |
| GSNETBackbone=ViT-B, Type=OVRSIS2026.04 | 75.98 | 84.51 | |
| CAT-SEGBackbone=ViT-L, Type=OVS2026.04 | 75.23 | 92.15 | |
| CAT-SEGBackbone=ViT-B, Type=OVS2026.04 | 74.84 | 81.51 | |
| Pi-SegBackbone=ViT-B, Type=OVRSIS2026.04 | 74.51 | 79.67 | |
| SANBackbone=ViT-L, Type=OVS2026.04 | 71.55 | 76.23 | |
| OVRSBackbone=ViT-L, Type=OVRSIS2026.04 | 70.99 | 76.46 | |
| OVRSBackbone=ViT-B, Type=OVRSIS2026.04 | 70.61 | 74.92 | |
| SCANBackbone=ViT-L, Type=OVS2026.04 | 69.39 | 79.3 | |
| SEDBackbone=ConvNeXt-B, Type=OVS2026.04 | 69.27 | 79.81 | |
| RSKT-SEGBackbone=ViT-B, Type=OVRSIS2026.04 | 68.98 | 77.25 | |
| SEDBackbone=ConvNeXt-L, Type=OVS2026.04 | 68.23 | 83.72 | |
| FCCLIPBackbone=ConvNeXt-L, Type=OVS2026.04 | 68.03 | 85.89 | |
| SANBackbone=ViT-B, Type=OVS2026.04 | 67.6 | 85.53 | |
| EOV-SEGBackbone=ConvNeXt-L, Type=OVS2026.04 | 67.37 | 80.98 | |
| FCCLIPBackbone=ConvNeXt-B, Type=OVS2026.04 | 65.89 | 79.07 | |
| EOV-SEGBackbone=ConvNeXt-B, Type=OVS2026.04 | 65.24 | 78.75 | |
| SCANBackbone=ViT-B, Type=OVS2026.04 | 63.82 | 84.74 | |
| RSKT-SEGBackbone=ViT-L, Type=OVRSIS2026.04 | 63.57 | 87.12 | |
| SegEarth-OVResolution=896 x 8962024.10 | 43.1 | — | |
| SegEarth-OVTraining Protocol=Trained on remote sensing data, Input Resolution=896x8962026.02 | 43.1 | — | |
| ConInfer2026.03 | 41.34 | — | |
| CLIP-SAMTraining Protocol=Zero-shot, Input Resolution=896x8962026.02 | 39.1 | — | |
| SegEarth-OVResolution=448 x 4482024.10 | 37 | — | |
| SegEarth-OVTraining/Evaluation Protocol=Trained on remote sensing data2026.02 | 37 | — | |
| SegEarth-OVBackbone=ViT-B, Type=OVRSIS (TF)2026.04 | 37 | — | |
| SegEarth-OV2026.03 | 35.95 | — | |
| OursTraining/Evaluation Protocol=Zero-shot2026.02 | 34.4 | — | |
| GEM2026.03 | 33.48 | — | |
| MaskCLIP*configuration=baseline2026.03 | 31.4 | — | |
| ClearCLIPResolution=448 x 4482024.10 | 30.1 | — | |
| ClearCLIPTraining/Evaluation Protocol=Zero-shot2026.02 | 30.1 | — | |
| ClearCLIPBackbone=ViT-B, Type=OVS (TF)2026.04 | 30.1 | — | |
| ClearCLIP2026.03 | 29.83 | — | |
| SCLIP2026.03 | 29.42 | — | |
| MaskCLIPResolution=448 x 4482024.10 | 29.2 | — | |
| MaskCLIPTraining/Evaluation Protocol=Zero-shot2026.02 | 29.2 | — | |
| MaskCLIPBackbone=ViT-B, Type=OVS (TF)2026.04 | 29.2 | — | |
| MaskCLIP2026.03 | 26.92 | — | |
| SCLIPResolution=448 x 4482024.10 | 25.9 | — | |
| SCLIPTraining/Evaluation Protocol=Zero-shot2026.02 | 25.9 | — | |
| SCLIPBackbone=ViT-B, Type=OVS (TF)2026.04 | 25.9 | — | |
| GEMResolution=448 x 4482024.10 | 20.8 | — | |
| GEMTraining/Evaluation Protocol=Zero-shot2026.02 | 20.8 | — | |
| GEMBackbone=ViT-B, Type=OVS (TF)2026.04 | 20.8 | — | |
| CLIPResolution=448 x 4482024.10 | 16 | — | |
| CLIPTraining/Evaluation Protocol=Zero-shot2026.02 | 16 | — | |
| CLIPBackbone=ViT-B, Type=OVS (TF)2026.04 | 16 | — | |
| CLIP2026.03 | 8.88 | — |