Segmentation on Vessels Many-shot
92Total DiceSwin
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SwinPre-training Strategy=Image+text pretraining, Backbone=Swin2026.05 | 92 | 87 | |
| Swin (over)Pre-training Strategy=Image+text pretraining, Backbone=Swin, Method Variant=overfit variant2026.05 | 92 | 88 | |
| UNetPre-training Strategy=Image+text pretraining, Backbone=UNet2026.05 | 92 | 81 | |
| uSAM (b)Pre-training Strategy=Baselines, Backbone=uSAM, Method Variant=base2026.05 | 90 | 81 | |
| UNetPre-training Strategy=Trained from scratch, Backbone=UNet2026.05 | 89 | 68 | |
| SwinPre-training Strategy=Image-only pretraining, Backbone=Swin2026.05 | 89 | 81 | |
| UNetPre-training Strategy=Image-only pretraining, Backbone=UNet2026.05 | 89 | 84 | |
| SwinPre-training Strategy=Trained from scratch, Backbone=Swin2026.05 | 88 | 76 | |
| uSAM (l)Pre-training Strategy=Baselines, Backbone=uSAM, Method Variant=large2026.05 | 84 | 82 |