Semantic Segmentation on ISIC (test)
2,341mIoUDistillFSS
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DistillFSSBackbone=Swin-B, Number of support images (K)=602025.12 | 2,341 | — | |
| TransferFSSBackbone=Swin-B, Number of support images (K)=602025.12 | 2,215 | — | |
| TransferFSSBackbone=Swin-B, Number of support images (K)=302025.12 | 2,048 | — | |
| DistillFSSBackbone=Swin-B, Number of support images (K)=302025.12 | 1,932 | — | |
| DMTNetBackbone=ResNet-50, Number of support images (K)=602025.12 | 1,735 | — | |
| PATNetBackbone=ResNet-50, Number of support images (K)=602025.12 | 1,697 | — | |
| PATNetBackbone=ResNet-50, Number of support images (K)=92025.12 | 1,601 | — | |
| PATNetBackbone=ResNet-50, Number of support images (K)=302025.12 | 1,595 | — | |
| DistillFSSBackbone=Swin-B, Number of support images (K)=152025.12 | 1,595 | — | |
| TransferFSSBackbone=Swin-B, Number of support images (K)=152025.12 | 1,581 | — | |
| TransferFSSBackbone=ResNet-50, Number of support images (K)=152025.12 | 1,565 | — | |
| DMTNetBackbone=ResNet-50, Number of support images (K)=302025.12 | 1,555 | — | |
| DMTNetBackbone=ResNet-50, Number of support images (K)=152025.12 | 1,507 | — | |
| DMTNetBackbone=ResNet-50, Number of support images (K)=92025.12 | 1,500 | — | |
| TransferFSSBackbone=ResNet-50, Number of support images (K)=602025.12 | 1,500 | — | |
| TransferFSSBackbone=Swin-B, Number of support images (K)=92025.12 | 1,435 | — | |
| TransferFSSBackbone=ResNet-50, Number of support images (K)=302025.12 | 1,434 | — | |
| LabelAnythingBackbone=ViT-B, Number of support images (K)=602025.12 | 1,390 | — | |
| PATNetBackbone=ResNet-50, Number of support images (K)=152025.12 | 1,346 | — | |
| DistillFSSBackbone=Swin-B, Number of support images (K)=92025.12 | 1,331 | — | |
| DistillFSSBackbone=ResNet-50, Number of support images (K)=152025.12 | 1,248 | — | |
| DistillFSSBackbone=ResNet-50, Number of support images (K)=302025.12 | 1,236 | — | |
| DistillFSSBackbone=ResNet-50, Number of support images (K)=602025.12 | 1,222 | — | |
| LabelAnythingBackbone=ViT-B, Number of support images (K)=302025.12 | 1,139 | — | |
| DistillFSSBackbone=ResNet-50, Number of support images (K)=92025.12 | 1,124 | — | |
| BAMBackbone=ResNet-50, Number of support images (K)=92025.12 | 967 | — | |
| TransferFSSBackbone=ResNet-50, Number of support images (K)=92025.12 | 938 | — | |
| HDMNetBackbone=ResNet-50, Number of support images (K)=92025.12 | 916 | — | |
| DCAMABackbone=ResNet-50, Number of support images (K)=152025.12 | 880 | — | |
| BAMBackbone=ResNet-50, Number of support images (K)=602025.12 | 869 | — | |
| DCAMABackbone=Swin-B, Number of support images (K)=92025.12 | 867 | — | |
| BAMBackbone=ResNet-50, Number of support images (K)=302025.12 | 847 | — | |
| BAMBackbone=ResNet-50, Number of support images (K)=152025.12 | 834 | — | |
| HDMNetBackbone=ResNet-50, Number of support images (K)=302025.12 | 820 | — | |
| HDMNetBackbone=ResNet-50, Number of support images (K)=152025.12 | 782 | — | |
| HDMNetBackbone=ResNet-50, Number of support images (K)=602025.12 | 777 | — | |
| DCAMABackbone=Swin-B, Number of support images (K)=602025.12 | 772 | — | |
| LabelAnythingBackbone=ViT-B, Number of support images (K)=152025.12 | 765 | — | |
| DCAMABackbone=ResNet-50, Number of support images (K)=602025.12 | 759 | — | |
| DCAMABackbone=Swin-B, Number of support images (K)=152025.12 | 725 | — | |
| DCAMABackbone=Swin-B, Number of support images (K)=302025.12 | 722 | — | |
| LabelAnythingBackbone=ViT-B, Number of support images (K)=92025.12 | 720 | — | |
| DCAMABackbone=ResNet-50, Number of support images (K)=302025.12 | 697 | — | |
| DCAMABackbone=ResNet-50, Number of support images (K)=92025.12 | 336 | — | |
| DeCon-SLPre-training Dataset=Downstream (ISIC), Label Percentage=100%, Transfer=enc, Loss=Lenc, Ldec2025.03 | 83.66 | — | |
| DeCon-SLPre-training Dataset=Downstream (ISIC), Label Percentage=100%, Transfer=enc + dec, Loss=Lenc, Ldec2025.03 | 83.25 | — | |
| SlotConPre-training Dataset=Downstream (ISIC), Label Percentage=100%, Transfer=enc, Loss=Lenc2025.03 | 83.19 | — | |
| DeCon-SLPre-training Dataset=COCO, Label Percentage=100%, Transfer=enc + dec, Loss=Lenc, Ldec2025.03 | 82.94 | — | |
| DeCon-SLPre-training Dataset=COCO, Label Percentage=100%, Transfer=enc, Loss=Lenc, Ldec2025.03 | 82.84 | — | |
| SlotConPre-training Dataset=COCO, Label Percentage=100%, Transfer=enc, Loss=Lenc2025.03 | 82.52 | — | |
| Random init.Pre-training Dataset=None, Label Percentage=100%2025.03 | 80.65 | — | |
| DeCon-SLPre-training Dataset=COCO, Label Percentage=25%, Transfer=enc + dec, Loss=Lenc, Ldec2025.03 | 79.97 | — | |
| DeCon-SLPre-training Dataset=COCO, Label Percentage=25%, Transfer=enc, Loss=Lenc, Ldec2025.03 | 79.56 | — | |
| SlotConPre-training Dataset=COCO, Label Percentage=25%, Transfer=enc, Loss=Lenc2025.03 | 79.05 | — | |
| Random init.Pre-training Dataset=None, Label Percentage=25%2025.03 | 78.38 | — | |
| DeCon-SLPre-training Dataset=COCO, Label Percentage=5%, Transfer=enc + dec, Loss=Lenc, Ldec2025.03 | 76 | — | |
| Random init.Pre-training Dataset=None, Label Percentage=5%2025.03 | 75.66 | — | |
| DeCon-SLPre-training Dataset=COCO, Label Percentage=5%, Transfer=enc, Loss=Lenc, Ldec2025.03 | 75.62 | — | |
| SlotConPre-training Dataset=COCO, Label Percentage=5%, Transfer=enc, Loss=Lenc2025.03 | 74.95 | — | |
| BaselineNoise Condition=Original data2021.07 | — | 0.8249 | |
| BaselineNoise ratio (alpha)=0.3, Noise level (beta)=0.52021.07 | — | 0.8075 | |
| BaselineNoise ratio (alpha)=0.3, Noise level (beta)=0.72021.07 | — | 0.7946 | |
| BaselineNoise ratio (alpha)=0.5, Noise level (beta)=0.52021.07 | — | 0.7895 | |
| BaselineNoise ratio (alpha)=0.5, Noise level (beta)=0.72021.07 | — | 0.7544 | |
| BaselineNoise ratio (alpha)=0.7, Noise level (beta)=0.52021.07 | — | 0.7661 | |
| BaselineNoise ratio (alpha)=0.7, Noise level (beta)=0.72021.07 | — | 0.7151 | |
| BaselineNoise ratio (alpha)=1.0, Noise level (beta)=0.52021.07 | — | 0.7113 | |
| BaselineNoise ratio (alpha)=1.0, Noise level (beta)=0.72021.07 | — | 0.6371 | |
| Co-teachingNoise Condition=Original data2021.07 | — | 0.8272 | |
| Co-teachingNoise ratio (alpha)=0.3, Noise level (beta)=0.52021.07 | — | 0.8144 | |
| Co-teachingNoise ratio (alpha)=0.3, Noise level (beta)=0.72021.07 | — | 0.8147 | |
| Co-teachingNoise ratio (alpha)=0.5, Noise level (beta)=0.52021.07 | — | 0.8122 | |
| Co-teachingNoise ratio (alpha)=0.5, Noise level (beta)=0.72021.07 | — | 0.8006 | |
| Co-teachingNoise ratio (alpha)=0.7, Noise level (beta)=0.52021.07 | — | 0.7961 | |
| Co-teachingNoise ratio (alpha)=0.7, Noise level (beta)=0.72021.07 | — | 0.785 | |
| Co-teachingNoise ratio (alpha)=1.0, Noise level (beta)=0.52021.07 | — | 0.7669 | |
| Co-teachingNoise ratio (alpha)=1.0, Noise level (beta)=0.72021.07 | — | 0.7368 | |
| JoCoRNoise Condition=Original data2021.07 | — | 0.8364 | |
| JoCoRNoise ratio (alpha)=0.3, Noise level (beta)=0.52021.07 | — | 0.8265 | |
| JoCoRNoise ratio (alpha)=0.3, Noise level (beta)=0.72021.07 | — | 0.8158 | |
| JoCoRNoise ratio (alpha)=0.5, Noise level (beta)=0.52021.07 | — | 0.8241 | |
| JoCoRNoise ratio (alpha)=0.5, Noise level (beta)=0.72021.07 | — | 0.8106 | |
| JoCoRNoise ratio (alpha)=0.7, Noise level (beta)=0.52021.07 | — | 0.8055 | |
| JoCoRNoise ratio (alpha)=0.7, Noise level (beta)=0.72021.07 | — | 0.7905 | |
| JoCoRNoise ratio (alpha)=1.0, Noise level (beta)=0.52021.07 | — | 0.7843 | |
| JoCoRNoise ratio (alpha)=1.0, Noise level (beta)=0.72021.07 | — | 0.743 | |
| Superpixel-guided Iterative LearningNoise Condition=Original data2021.07 | — | 0.8426 | |
| Superpixel-guided Iterative LearningNoise ratio (alpha)=0.3, Noise level (beta)=0.52021.07 | — | 0.84 | |
| Superpixel-guided Iterative LearningNoise ratio (alpha)=0.3, Noise level (beta)=0.72021.07 | — | 0.8334 | |
| Superpixel-guided Iterative LearningNoise ratio (alpha)=0.5, Noise level (beta)=0.52021.07 | — | 0.839 | |
| Superpixel-guided Iterative LearningNoise ratio (alpha)=0.5, Noise level (beta)=0.72021.07 | — | 0.8319 | |
| Superpixel-guided Iterative LearningNoise ratio (alpha)=0.7, Noise level (beta)=0.52021.07 | — | 0.8383 | |
| Superpixel-guided Iterative LearningNoise ratio (alpha)=0.7, Noise level (beta)=0.72021.07 | — | 0.8312 | |
| Superpixel-guided Iterative LearningNoise ratio (alpha)=1.0, Noise level (beta)=0.52021.07 | — | 0.8223 | |
| Superpixel-guided Iterative LearningNoise ratio (alpha)=1.0, Noise level (beta)=0.72021.07 | — | 0.8139 | |
| Tri-networkNoise Condition=Original data2021.07 | — | 0.8296 | |
| Tri-networkNoise ratio (alpha)=0.3, Noise level (beta)=0.52021.07 | — | 0.815 | |
| Tri-networkNoise ratio (alpha)=0.3, Noise level (beta)=0.72021.07 | — | 0.8073 | |
| Tri-networkNoise ratio (alpha)=0.5, Noise level (beta)=0.52021.07 | — | 0.8094 | |
| Tri-networkNoise ratio (alpha)=0.5, Noise level (beta)=0.72021.07 | — | 0.8024 |