Few-shot segmentation on FSS-1000
86.51mIoU (1-shot)GF-SAM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GF-SAMBackbone=DINOv3-B+SAM-B, Training Approach=Training-free, Task=FSS, Mark=NIPS-242026.06 | 86.51 | — | 87.57 | |
| GF-SAMBackbone=DINOv2-B+SAM-B, Training Approach=Training-free, Task=FSS, Mark=NIPS-242026.06 | 85.75 | — | 86.81 | |
| MatcherBackbone=DINOv2-B+SAM-B, Training Approach=Training-free, Task=FSS, Mark=ICLR-242026.06 | 85.27 | — | 88.52 | |
| LoECBackbone=DINOv3-B, Training Approach=Training-based, Task=CD-FSS, Mark=CVPR-252026.06 | 84.33 | — | 86.16 | |
| PATNetBackbone=DINOv3-B, Training Approach=Training-based, Task=CD-FSS, Mark=ECCV-222026.06 | 83.71 | — | 84.65 | |
| MatcherBackbone=DINOv3-B+SAM-B, Training Approach=Training-free, Task=FSS, Mark=ICLR-242026.06 | 83.39 | — | 87.44 | |
| DFNBackbone=ViT-B, Training Approach=Training-based, Task=CD-FSS, Mark=ICML-252026.06 | 82.97 | — | 85.72 | |
| OursBackbone=DINOv3-B, Training Approach=Training-free, Task=CD-FSS, Mark=Ours2026.06 | 82.67 | — | 83.89 | |
| LoECBackbone=ViT-B, Training Approach=Training-based, Task=CD-FSS, Mark=CVPR-252026.06 | 81.05 | — | 83.69 | |
| FPTransBackbone=ViT-B, Training Approach=Training-based, Task=FSS, Mark=NIPS-222026.06 | 80.74 | — | 83.65 | |
| FSSAMBackbone=DINOv2-B+SAM2, Training Approach=Training-based, Task=FSS, Mark=ICML-252026.06 | 80.36 | — | 87.11 | |
| SDRCBackbone=ViT-B, Training Approach=Training-based, Task=CD-FSS, Mark=ICML-252026.06 | 80.31 | — | 82.55 | |
| APSegBackbone=SAM-B, Training Approach=Training-based, Task=CD-FSS, Mark=CVPR-242026.06 | 79.71 | — | 81.9 | |
| APMBackbone=ResNet-50, Training Approach=Training-based, Task=CD-FSS, Mark=NIPS-242026.06 | 79.29 | — | 81.83 | |
| DRABackbone=ResNet-50, Training Approach=Training-based, Task=CD-FSS, Mark=CVPR-242026.06 | 79.05 | — | 80.4 | |
| SSPBackbone=ResNet-50, Training Approach=Training-based, Task=FSS, Mark=ECCV-222026.06 | 78.91 | — | 80.59 | |
| ISABackbone=ResNet-50, Training Approach=Training-based, Task=CD-FSS, Mark=ICCV-252026.06 | 78.8 | — | 86 | |
| PATNetBackbone=ResNet-50, Training Approach=Training-based, Task=CD-FSS, Mark=ECCV-222026.06 | 78.59 | — | 81.23 | |
| ABCDFSSBackbone=DINOv3-B, Training Approach=Training-based, Task=CD-FSS, Mark=CVPR-242026.06 | 76.8 | — | 79.39 | |
| HDMNetBackbone=ResNet-50, Training Approach=Training-based, Task=FSS, Mark=CVPR-232026.06 | 75.1 | — | 78.6 | |
| ABCDFSSBackbone=ResNet-50, Training Approach=Training-based, Task=CD-FSS, Mark=CVPR-242026.06 | 74.6 | — | 76.2 | |
| MAUPBackbone=DINOv2-B+SAM-B, Training Approach=Training-free, Task=FSMIS, Mark=MICCAI-252026.06 | 73.29 | — | 75.97 | |
| PerSAMBackbone=SAM-B, Training Approach=Training-free, Task=FSS, Mark=ICLR-242026.06 | 60.92 | — | 66.53 | |
| ABCDFSSBackbone=ResNet-50, Publication=CVPR 2024, Shot=1-shot2025.11 | — | 74.6 | — | |
| ABCDFSSBackbone=ResNet-50, Publication=CVPR 2024, Shot=5-shot2025.11 | — | 76.2 | — | |
| APMBackbone=ResNet-50, Publication=NeurIPS 2024, Shot=1-shot2025.11 | — | 79.3 | — | |
| APMBackbone=ResNet-50, Publication=NeurIPS 2024, Shot=5-shot2025.11 | — | 81.9 | — | |
| APSegBackbone=Vit-base, Publication=CVPR 2024, Shot=1-shot2025.11 | — | 79.7 | — | |
| APSegBackbone=Vit-base, Publication=CVPR 2024, Shot=5-shot2025.11 | — | 81.9 | — | |
| DCDNetBackbone=ResNet-50, Publication=-, Shot=1-shot2025.11 | — | 81.7 | — | |
| DCDNetBackbone=ResNet-50, Publication=-, Shot=5-shot2025.11 | — | 83.3 | — | |
| DR-AdapterBackbone=ResNet-50, Publication=CVPR 2024, Shot=1-shot2025.11 | — | 79.1 | — | |
| DR-AdapterBackbone=ResNet-50, Publication=CVPR 2024, Shot=5-shot2025.11 | — | 80.4 | — | |
| HSNetBackbone=ResNet-50, Publication=ICCV 2021, Shot=1-shot2025.11 | — | 77.5 | — | |
| HSNetBackbone=ResNet-50, Publication=ICCV 2021, Shot=5-shot2025.11 | — | 81 | — | |
| IFABackbone=ResNet-50, Publication=CVPR 2024, Shot=1-shot2025.11 | — | 80.1 | — | |
| IFABackbone=ResNet-50, Publication=CVPR 2024, Shot=5-shot2025.11 | — | 82.4 | — | |
| LoECBackbone=ResNet-50, Publication=CVPR 2025, Shot=1-shot2025.11 | — | 78.5 | — | |
| LoECBackbone=ResNet-50, Publication=CVPR 2025, Shot=5-shot2025.11 | — | 80.6 | — | |
| PATNetBackbone=ResNet-50, Publication=CVPR 2022, Shot=1-shot2025.11 | — | 78.6 | — | |
| PATNetBackbone=ResNet-50, Publication=CVPR 2022, Shot=5-shot2025.11 | — | 81.2 | — | |
| PerSAMBackbone=Vit-base, Publication=ICLR 2023, Shot=1-shot2025.11 | — | 60.9 | — | |
| PerSAMBackbone=Vit-base, Publication=ICLR 2023, Shot=5-shot2025.11 | — | 66.5 | — | |
| SSPBackbone=ResNet-50, Publication=ECCV 2022, Shot=1-shot2025.11 | — | 78.9 | — | |
| SSPBackbone=ResNet-50, Publication=ECCV 2022, Shot=5-shot2025.11 | — | 80.6 | — |