Semantic Segmentation on LoveDA
78.2mIoUGSNet
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GSNetRS Training=Yes, In-domain supervised training=true2026.05 | 78.2 | — | |
| CAFe-DINORS Training=No2026.05 | 65.3 | — | |
| Co2SLabeled Ratio=1/42025.12 | 64 | — | |
| DWLLabeled Ratio=1/42025.12 | 63.4 | — | |
| MUCALabeled Ratio=1/42025.12 | 63.1 | — | |
| DWLLabeled Ratio=1/82025.12 | 62.8 | — | |
| Co2SLabeled Ratio=1/82025.12 | 62.7 | — | |
| UniMatchLabeled Ratio=1/42025.12 | 61.7 | — | |
| FixMatchLabeled Ratio=1/42025.12 | 61.5 | — | |
| U2PLLabeled Ratio=1/42025.12 | 61.5 | — | |
| WSCLLabeled Ratio=1/42025.12 | 61.1 | — | |
| Co2SLabeled Ratio=1/162025.12 | 60.4 | — | |
| MUCALabeled Ratio=1/82025.12 | 60.2 | — | |
| FixMatchLabeled Ratio=1/82025.12 | 60.1 | — | |
| UniMatchLabeled Ratio=1/82025.12 | 60 | — | |
| U2PLLabeled Ratio=1/82025.12 | 59.6 | — | |
| WSCLLabeled Ratio=1/82025.12 | 59.4 | — | |
| DWLLabeled Ratio=1/162025.12 | 59 | — | |
| OnlySupLabeled Ratio=1/42025.12 | 58.9 | — | |
| UniMatchLabeled Ratio=1/162025.12 | 58.6 | — | |
| Co2SLabeled Ratio=1/402025.12 | 58.2 | — | |
| U2PLLabeled Ratio=1/162025.12 | 58 | — | |
| FixMatchLabeled Ratio=1/162025.12 | 57.9 | — | |
| WSCLLabeled Ratio=1/162025.12 | 57.4 | — | |
| DWLLabeled Ratio=1/402025.12 | 57.4 | — | |
| MUCALabeled Ratio=1/162025.12 | 56.7 | — | |
| OnlySupLabeled Ratio=1/82025.12 | 56.6 | — | |
| UniMatchLabeled Ratio=1/402025.12 | 55.5 | — | |
| CoDe-MAEBackbone=HiViT-B2026.04 | 55.32 | — | |
| DINOv3-H+Evaluation Head=DPT Head2025.12 | 55.3 | — | |
| DINOv2-LEvaluation Head=DPT Head2025.12 | 55.2 | — | |
| DINOv2-gEvaluation Head=DPT Head2025.12 | 55 | — | |
| Pixio-HEvaluation Head=DPT Head2025.12 | 54.7 | — | |
| MUCALabeled Ratio=1/402025.12 | 54.6 | — | |
| BillionFMBackbone=ViT-G12X4, Decoder=UperNet2024.03 | 54.4 | — | |
| SelectiveMAEBackbone=ViT-L [2], Params (M)=307, Throughput/Minute=533k, Pre-training=OpticalRS-13 (4M)2024.06 | 54.31 | — | |
| SelectiveMAEBackbone=ViT-L2026.04 | 54.31 | — | |
| MAE + MTPBackbone=ViT-L + RVSA, Decoder=UperNet2024.03 | 54.17 | — | |
| IMP + MTPBackbone=InternImage-XL, Decoder=UperNet2024.03 | 54.17 | — | |
| AerialFormerBackbone=Swin-B, Decoder=UperNet2024.03 | 54.1 | — | |
| WSCLLabeled Ratio=1/402025.12 | 54.1 | — | |
| OREOLEBackbone=VIT-G [43], Params (M)=9142024.06 | 54 | — | |
| LSKNetBackbone=LSKNet-S, Decoder=UperNet2024.03 | 54 | — | |
| IMPBackbone=InternImage-XL, Decoder=UperNet2024.03 | 53.93 | — | |
| SelectiveMAEBackbone=ViT-L [2], Params (M)=307, Throughput/Minute=533k2024.06 | 53.92 | — | |
| Pixio-HEvaluation Head=Linear Head2025.12 | 53.9 | — | |
| MAEBackbone=ViT-L + RVSA, Decoder=UperNet2024.03 | 53.72 | — | |
| SegEarth-OVRS Training=Self-Supervised2026.05 | 53.5 | — | |
| FixMatchLabeled Ratio=1/402025.12 | 53.3 | — | |
| OnlySupLabeled Ratio=1/162025.12 | 53.1 | — | |
| SelectiveMAEBackbone=ViT-B [2], Params (M)=86, Throughput/Minute=556k, Pre-training=OpticalRS-13 (4M)2024.06 | 53.05 | — | |
| SelectiveMAEBackbone=ViT-B2026.04 | 53.05 | — | |
| SMLFRBackbone=ConvNext-L, Decoder=UperNet2024.03 | 53.03 | — | |
| UperNetPretrain=RSP, Backbone=VITAEv2-S2022.08 | 53.02 | — | |
| RSPBackbone=VITAEv2-S, Decoder=UperNet2024.03 | 53.02 | — | |
| MAEBackbone=ViT-B [2], Params (M)=86, Throughput/Minute=264k, Pre-training=OpticalRS-13 (4M)2024.06 | 52.8 | — | |
| DINOv3-H+Evaluation Head=Linear Head2025.12 | 52.7 | — | |
| SelectiveMAEBackbone=ViT-B [2], Params (M)=86, Throughput/Minute=556k2024.06 | 52.68 | — | |
| UperNetPretrain=MAE, Backbone=VITAE-B2022.08 | 52.58 | — | |
| Hi-ResNetBackbone=Hi-ResNet, Decoder=UperNet2024.03 | 52.5 | — | |
| UperNetPretrain=MAE, Backbone=VITAE-B + RVSA*2022.08 | 52.44 | — | |
| UperNetBackbone=ViTAE-B + RVSA, Decoder=UperNet2024.03 | 52.44 | — | |
| RSSFormerBackbone=RSS-B, Decoder=UperNet2024.03 | 52.43 | — | |
| UNetFormerBackbone=ResNet-18, Decoder=UperNet2024.03 | 52.4 | — | |
| MAE + MTPBackbone=ViT-B + RVSA, Decoder=UperNet2024.03 | 52.39 | — | |
| U2PLLabeled Ratio=1/402025.12 | 52.3 | — | |
| UperNetPretrain=MAE, Backbone=VITAE-B + RVSA2022.08 | 52.26 | — | |
| MaRSBackbone=Swin V2-B2026.04 | 52.25 | — | |
| DINOv2-LEvaluation Head=Linear Head2025.12 | 52.2 | — | |
| RVSABackbone=VIT-B+RVSA [21], Params (M)=862024.06 | 51.95 | — | |
| UperNetPretrain=MAE, Backbone=ViT-B + RVSA2022.08 | 51.95 | — | |
| UperNetPretrain=MAE, Backbone=VIT-B + RVSA*2022.08 | 51.95 | — | |
| MAEBackbone=ViT-B + RVSA, Decoder=UperNet2024.03 | 51.95 | — | |
| RVSABackbone=ViT-B+RVSA2026.04 | 51.95 | — | |
| DINOv2-gEvaluation Head=Linear Head2025.12 | 51.9 | — | |
| UperNetPretrain=MAE, Backbone=VITAE-B + VSA2022.08 | 51.82 | — | |
| UperNetPretrain=MAE, Backbone=ViT-B + VSA2022.08 | 51.29 | — | |
| UperNetPretrain=IMP, Backbone=ResNet-502022.08 | 51.27 | — | |
| UperNetPretrain=MAE, Backbone=ViT-B2022.08 | 51.03 | — | |
| DCFAMBackbone=Swin-T, Decoder=UperNet2024.03 | 50.6 | — | |
| GeRSPBackbone=ResNet-50, Decoder=UperNet2024.03 | 50.56 | — | |
| MAE-HEvaluation Head=DPT Head2025.12 | 50.2 | — | |
| UperNetPretrain=IMP, Backbone=Swin-T2022.08 | 50 | — | |
| OracleTraining Regime=Oracle (fully supervised SegFormer-b0)2025.12 | 50 | — | |
| HRNetBackbone=HRNet-W32, Decoder=UperNet2024.03 | 49.79 | — | |
| TOVBackbone=ResNet-50 [101], Params (M)=262024.06 | 49.7 | — | |
| TOVBackbone=ResNet-50, Decoder=UperNet2024.03 | 49.7 | — | |
| FactSegPretrain=IMP, Backbone=ResNet-502022.08 | 48.94 | — | |
| FactSegBackbone=ResNet-50, Decoder=UperNet2024.03 | 48.94 | — | |
| FSS-TIsShot=5-shot, Source Domain=DeepGlobe2025.09 | 48.92 | 3.09 | |
| CACOBackbone=ResNet-50 [101], Params (M)=262024.06 | 48.89 | — | |
| CACOBackbone=ResNet-50, Decoder=UperNet2024.03 | 48.89 | — | |
| GASSLBackbone=ResNet-50 [101], Params (M)=262024.06 | 48.76 | — | |
| GASSLBackbone=ResNet-50, Decoder=UperNet2024.03 | 48.76 | — | |
| PSPNetPretrain=IMP, Backbone=ResNet-502022.08 | 48.31 | — | |
| DeeplabV3+Pretrain=IMP, Backbone=ResNet-502022.08 | 48.31 | — | |
| PSPNetBackbone=ResNet-50, Decoder=UperNet2024.03 | 48.31 | — | |
| DeeplabV3+Backbone=ResNet-50, Decoder=UperNet2024.03 | 48.31 | — | |
| Semantic FPNPretrain=IMP, Backbone=ResNet-502022.08 | 48.15 | — | |
| FarSegPretrain=IMP, Backbone=ResNet-502022.08 | 48.15 | — |