Few-shot Semantic Segmentation on PASCAL-5^i (1-shot)
81.2mIoUFSS-SAM3 + text
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| FSS-SAM3 + textMethod Category=Foundation-based FSS Methods, Shot Count=1-shot, Backbone=SAM3, Resolution=1008 x 1008, Textual Guidance=true2026.04 | 81.2 | 76.2 | 87.5 | 81.6 | 79.5 | — | — | 89.4 | |
| FSSAMMethod Category=Foundation-based FSS Methods, Shot Count=1-shot2026.04 | 81 | 81.6 | 84.9 | 81.6 | 76 | — | — | 89.4 | |
| FSS-SAM3Method Category=Foundation-based FSS Methods, Shot Count=1-shot, Backbone=SAM3, Resolution=1008 x 1008, Textual Guidance=false2026.04 | 79.6 | 70.4 | 84.6 | 80.9 | 82.3 | — | — | 88.3 | |
| GF-SAMMethod Category=Foundation-based FSS Methods, Shot Count=1-shot2026.04 | 72.1 | 71.1 | 75.7 | 69.2 | 73.3 | — | — | — | |
| VRP-SAMMethod Category=Foundation-based FSS Methods, Shot Count=1-shot2026.04 | 71.9 | 73.9 | 78.3 | 70.6 | 65.1 | — | — | — | |
| HMNetMethod Category=Classical FSS Methods, Shot Count=1-shot2026.04 | 70.4 | 72.2 | 75.4 | 70 | 63.9 | — | — | 81.6 | |
| AMNetMethod Category=Classical FSS Methods, Shot Count=1-shot2026.04 | 70.1 | 71.1 | 75.9 | 69.7 | 63.7 | — | — | — | |
| AENetMethod Category=Classical FSS Methods, Shot Count=1-shot2026.04 | 69.8 | 72.2 | 75.5 | 68.5 | 63.1 | — | — | 80.8 | |
| PAMMethod Category=Classical FSS Methods, Shot Count=1-shot2026.04 | 69.5 | 71.1 | 75.5 | 67 | 64.5 | — | — | — | |
| HDMNetMethod Category=Classical FSS Methods, Shot Count=1-shot2026.04 | 69.4 | 71 | 75.4 | 68.9 | 62.1 | — | — | — | |
| FPTransMethod Category=Classical FSS Methods, Shot Count=1-shot2026.04 | 68.8 | 72.3 | 70.6 | 68.3 | 64.1 | — | — | — | |
| MatcherMethod Category=Foundation-based FSS Methods, Shot Count=1-shot2026.04 | 68.1 | 67.7 | 70.7 | 66.9 | 67 | — | — | — | |
| BAMMethod Category=Classical FSS Methods, Shot Count=1-shot2026.04 | 67.8 | 69 | 73.6 | 67.6 | 61.1 | — | — | 79.7 | |
| FECANetBackbone=ResNet-50, Setting=1-shot, Training time=33h, Inference time=122ms, Learnable params=3.5M2023.01 | 67.4 | — | — | — | — | — | — | — | |
| SCCANMethod Category=Classical FSS Methods, Shot Count=1-shot2026.04 | 66.8 | 68.3 | 72.5 | 66.8 | 59.8 | — | — | 77.7 | |
| HSNetBackbone=ResNet-50, Setting=1-shot, Training time=54h, Inference time=101ms, Learnable params=2.6M2023.01 | 64 | — | — | — | — | — | — | — | |
| HSNetBackbone=ResNet50, Annotation Type=Pixel2023.03 | 64 | 64.3 | 70.7 | 60.3 | 60.5 | — | — | — | |
| HSNetMethod Category=Classical FSS Methods, Shot Count=1-shot2026.04 | 64 | 64.3 | 70.7 | 60.3 | 60.5 | — | — | — | |
| MMNetBackbone=ResNet-50, Setting=1-shot, Training time=64h, Inference time=128ms, Learnable params=10.4M2023.01 | 61.8 | — | — | — | — | — | — | — | |
| SCLBackbone=ResNet50, Annotation Type=Pixel2023.03 | 61.8 | 63 | 70 | 56.5 | 57.7 | — | — | — | |
| IMR-HSNetBackbone=ResNet50, Annotation Type=Image2023.03 | 61.1 | 62.6 | 69.1 | 56.1 | 56.7 | — | — | — | |
| PFENetBackbone=ResNet-50, Setting=1-shot, Training time=24h, Inference time=52ms, Learnable params=10.8M2023.01 | 60.8 | — | — | — | — | — | — | — | |
| PFENetBackbone=ResNet50, Annotation Type=Pixel2023.03 | 60.8 | 61.7 | 69.5 | 55.4 | 56.3 | — | — | — | |
| PFENetMethod Category=Classical FSS Methods, Shot Count=1-shot2026.04 | 60.8 | 61.7 | 69.5 | 55.4 | 56.3 | — | — | — | |
| RePRIBackbone=ResNet-50, Setting=1-shot, Training time=17h, Inference time=72ms, Learnable params=46.7M2023.01 | 59.8 | — | — | — | — | — | — | — | |
| SS-PFENetBackbone=VGG16, Annotation Type=Pixel2023.03 | 59.8 | 54.5 | 67.4 | 63.4 | 54 | — | — | — | |
| HSNetBackbone=VGG16, Annotation Type=Pixel2023.03 | 59.7 | 59.6 | 65.7 | 59.6 | 54 | — | — | — | |
| RePRIBackbone=ResNet50, Annotation Type=Pixel2023.03 | 59.3 | 58.8 | 67.9 | 56.8 | 53.7 | — | — | — | |
| Ours (w/ trans.)Learning=Transductive, Backbone=ResNet-502024.04 | 58.11 | — | — | — | — | 76.39 | 39.83 | — | |
| ASRBackbone=ResNet50, Annotation Type=Pixel2023.03 | 56.9 | 53.8 | 69.6 | 51.6 | 52.8 | — | — | — | |
| IMR-HSNetBackbone=VGG16, Annotation Type=Image2023.03 | 56.5 | 58.2 | 63.9 | 52.9 | 51.2 | — | — | — | |
| CWTBackbone=ResNet-50, Setting=1-shot, Training time=10h, Inference time=232ms, Learnable params=2.1M2023.01 | 56.3 | — | — | — | — | — | — | — | |
| Ours (w/o trans.)Learning=Inductive, Backbone=ResNet-502024.04 | 54.79 | — | — | — | — | 74.58 | 34.99 | — | |
| POPLearning=Transductive, Backbone=ResNet-502024.04 | 54.72 | — | — | — | — | 73.92 | 35.51 | — | |
| ASRBackbone=VGG16, Annotation Type=Pixel2023.03 | 54.6 | 49.2 | 65.4 | 52.6 | 51.3 | — | — | — | |
| Siam et al. + CANetBackbone=ResNet50, Annotation Type=Image2023.03 | 53.5 | 49.5 | 65.5 | 50 | 49.2 | — | — | — | |
| RPMMBackbone=VGG16, Annotation Type=Pixel2023.03 | 53 | 47.1 | 65.8 | 50.6 | 48.5 | — | — | — | |
| DIaMLearning=Transductive, Backbone=ResNet-502024.04 | 53 | — | — | — | — | 70.89 | 35.11 | — | |
| SS-PANetBackbone=VGG16, Annotation Type=Pixel2023.03 | 52.3 | 49.3 | 60.8 | 53.9 | 45.2 | — | — | — | |
| CANetBackbone=ResNet50, Annotation Type=Box2023.03 | 52 | — | — | — | — | — | — | — | |
| Siam et al. + PANetBackbone=ResNet50, Annotation Type=Image2023.03 | 50.5 | 42.5 | 64.8 | 48.1 | 46.5 | — | — | — | |
| BAMLearning=Inductive, Backbone=ResNet-502024.04 | 49.55 | — | — | — | — | 71.6 | 27.49 | — | |
| DIaM (w/o trans.)Learning=Inductive, Backbone=ResNet-502024.04 | 47.08 | — | — | — | — | 66.79 | 27.36 | — | |
| PANetBackbone=VGG16, Annotation Type=Box2023.03 | 45.1 | — | — | — | — | — | — | — | |
| Lee et al.Backbone=VGG16, Annotation Type=Image2023.03 | 42.4 | 36.5 | 51.7 | 45.9 | 35.6 | — | — | — | |
| CAPLLearning=Inductive, Backbone=ResNet-502024.04 | 41.13 | — | — | — | — | 64.8 | 17.46 | — | |
| MiBLearning=Inductive, Backbone=ResNet-502024.04 | 36.33 | — | — | — | — | 63.8 | 8.86 | — | |
| POPLearning=Inductive, Backbone=ResNet-502024.04 | 33.32 | — | — | — | — | 46.68 | 19.96 | — | |
| PANETLearning=Inductive, Backbone=ResNet-502024.04 | 21.57 | — | — | — | — | 31.88 | 11.25 | — | |
| RePRILearning=Transductive, Backbone=ResNet-502024.04 | 15.63 | — | — | — | — | 20.76 | 10.5 | — | |
| SCLLearning=Inductive, Backbone=ResNet-502024.04 | 5.66 | — | — | — | — | 8.88 | 2.44 | — | |
| CANeTLearning=Inductive, Backbone=ResNet-502024.04 | 5.58 | — | — | — | — | 8.73 | 2.42 | — | |
| PFENETLearning=Inductive, Backbone=ResNet-502024.04 | 5.5 | — | — | — | — | 8.32 | 2.67 | — |