Medical Image Segmentation on SA-Med2D 20M (test)
88.2DSC (Bbox)SyncSAM-SAMed-50
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SyncSAM-SAMed-50Backbone=SAMed-502024.08 | 88.2 | 74.8 | 88.5 | |
| SyncSAM-SAMed-34Backbone=SAMed-342024.08 | 87 | 74.1 | 87.6 | |
| MedSAMTraining Strategy=Fully fine-tuned2024.08 | 80.8 | — | — | |
| SAM-Med2DTraining Strategy=Adapter-tuned2024.08 | 78.2 | 68.3 | 76.7 | |
| FT-SAMTraining Strategy=Mask decoder trained only2024.08 | 74.6 | 61.1 | 73.2 | |
| SAMTraining Strategy=Direct inference2024.08 | 66.6 | 24.5 | 52.8 |