Few-shot Segmentation on FSS-1000 (test)
91.7mIoUDACM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DACMBackbone=ResNet101, n-shot=5-shot2022.07 | 91.7 | — | — | |
| DACMBackbone=ResNet50, n-shot=5-shot2022.07 | 91.6 | — | — | |
| VAT + MSIBackbone=ResNet101, k-shot=5-shot2022.12 | 91 | — | — | |
| SANSAshot=5-shot, Training Protocol=Training-based, In-domain training=false2026.06 | 91 | — | — | |
| DACMBackbone=ResNet101, n-shot=1-shot2022.07 | 90.8 | — | — | |
| VATBackbone=ResNet-101, Shot=5-shot2022.07 | 90.8 | 94.4 | 68.6 | |
| DACMBackbone=ResNet50, n-shot=1-shot2022.07 | 90.7 | — | — | |
| VATBackbone=ResNet-50, Shot=5-shot2022.07 | 90.7 | 94.2 | 68.4 | |
| VATBackbone=ResNet101, n-shot=5-shot2022.07 | 90.6 | — | — | |
| VAT + MSIBackbone=ResNet50, k-shot=5-shot2022.12 | 90.6 | — | — | |
| VATBackbone=ResNet101, k-shot=5-shot2022.12 | 90.6 | — | — | |
| VAT + MSIBackbone=ResNet101, k-shot=1-shot2022.12 | 90.6 | — | — | |
| VATBackbone=ResNet50, n-shot=5-shot2022.07 | 90.3 | — | — | |
| VATBackbone=ResNet-101, Shot=1-shot2022.07 | 90.3 | 94 | 68 | |
| VATBackbone=ResNet50, k-shot=5-shot2022.12 | 90.3 | — | — | |
| CG-ICSshot=1-shot, Training Protocol=Training-free, In-domain training=false2026.06 | 90.2 | — | — | |
| VATBackbone=ResNet-50, Shot=1-shot2022.07 | 90.1 | 93.8 | 68.3 | |
| VATBackbone=ResNet101, n-shot=1-shot2022.07 | 90 | — | — | |
| VAT + MSIBackbone=ResNet50, k-shot=1-shot2022.12 | 90 | — | — | |
| VATBackbone=ResNet101, k-shot=1-shot2022.12 | 90 | — | — | |
| SANSAshot=1-shot, Training Protocol=Training-based, In-domain training=false2026.06 | 90 | — | — | |
| Matchershot=5-shot, Training Protocol=Training-free, In-domain training=false2026.06 | 89.6 | — | — | |
| VATBackbone=ResNet50, n-shot=1-shot2022.07 | 89.5 | — | — | |
| VATBackbone=ResNet50, k-shot=1-shot2022.12 | 89.5 | — | — | |
| SegGPTshot=5-shot, Training Protocol=Training-based, In-domain training=false2026.06 | 89.3 | — | — | |
| CG-ICSshot=5-shot, Training Protocol=Training-free, In-domain training=false2026.06 | 89.3 | — | — | |
| HSNet + MSIBackbone=ResNet101, k-shot=5-shot2022.12 | 89.2 | — | — | |
| GF-SAMshot=5-shot, Training Protocol=Training-free, In-domain training=false2026.06 | 88.9 | — | — | |
| HSNetBackbone=ResNet101, n-shot=5-shot2022.07 | 88.5 | — | — | |
| HSNetBackbone=ResNet-101, Shot=5-shot2022.07 | 88.5 | 92.9 | 63.6 | |
| HSNetBackbone=ResNet101, k-shot=5-shot2022.12 | 88.5 | — | — | |
| HSNet + MSIBackbone=ResNet50, k-shot=5-shot2022.12 | 88.4 | — | — | |
| DANBackbone=ResNet101, n-shot=5-shot2022.07 | 88.1 | — | — | |
| DANBackbone=ResNet-101, Shot=5-shot2022.07 | 88.1 | — | — | |
| HSNet + MSIBackbone=ResNet101, k-shot=1-shot2022.12 | 88.1 | — | — | |
| DiffwSshot=5-shot, Training Protocol=Training-based, In-domain training=false2026.06 | 88 | — | — | |
| GF-SAMshot=1-shot, Training Protocol=Training-free, In-domain training=false2026.06 | 88 | — | — | |
| HSNetBackbone=ResNet50, n-shot=5-shot2022.07 | 87.8 | — | — | |
| HSNetBackbone=ResNet-50, Shot=5-shot2022.07 | 87.8 | 92.5 | 63.3 | |
| HSNetBackbone=ResNet50, k-shot=5-shot2022.12 | 87.8 | — | — | |
| DiffwSshot=1-shot, Training Protocol=Training-based, In-domain training=false2026.06 | 87.8 | — | — | |
| HSNet + MSIBackbone=ResNet50, k-shot=1-shot2022.12 | 87.5 | — | — | |
| Matchershot=1-shot, Training Protocol=Training-free, In-domain training=false2026.06 | 87 | — | — | |
| HSNetBackbone=ResNet101, n-shot=1-shot2022.07 | 86.5 | — | — | |
| HSNetBackbone=ResNet-101, Shot=1-shot2022.07 | 86.5 | 91.6 | 62.4 | |
| HSNetBackbone=ResNet101, k-shot=1-shot2022.12 | 86.5 | — | — | |
| UNICL-SAMshot=5-shot, Training Protocol=Training-based, In-domain training=false2026.06 | 86.3 | — | — | |
| SegGPTshot=1-shot, Training Protocol=Training-based, In-domain training=false2026.06 | 85.6 | — | — | |
| HSNetBackbone=ResNet50, n-shot=1-shot2022.07 | 85.5 | — | — | |
| HSNetBackbone=ResNet-50, Shot=1-shot2022.07 | 85.5 | 91 | 62.1 | |
| HSNetBackbone=ResNet50, k-shot=1-shot2022.12 | 85.5 | — | — | |
| DANBackbone=ResNet101, n-shot=1-shot2022.07 | 85.2 | — | — | |
| DANBackbone=ResNet-101, Shot=1-shot2022.07 | 85.2 | — | — | |
| EfficientLab + PRNbase_network=EfficientLab2022.11 | 84.1 | — | 73.2 | |
| UNICL-SAMshot=1-shot, Training Protocol=Training-based, In-domain training=false2026.06 | 84 | — | — | |
| FSOTBackbone=ResNet50, n-shot=5-shot2022.07 | 83.8 | — | — | |
| FSOTBackbone=ResNet-50, Shot=5-shot2022.07 | 83.8 | — | — | |
| EfficientLab2022.11 | 82.8 | — | 71.1 | |
| Adapted Relation Network + PRNbase_network=Adapted Relation Network2022.11 | 82.7 | — | 72.9 | |
| FSOTBackbone=ResNet50, n-shot=1-shot2022.07 | 82.5 | — | — | |
| FSOTBackbone=ResNet-50, Shot=1-shot2022.07 | 82.5 | — | — | |
| Adapted Relation Network2022.11 | 80.1 | — | 69.8 | |
| Paintersetting=open-vocabulary2022.12 | 62.3 | — | — | |
| MAE-VQGANsetting=open-vocabulary2022.12 | 58.3 | — | — |