Layer-wise search-based merging of SAM, MedSAM, and MedicoSAM improves average Dice on 25 tasks by 6.67 points when tuned per task and 4.37 points when tuned across tasks.
The medical segmentation decathlon
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
MedSAMix: A Training-Free Model Merging Approach for Medical Image Segmentation
Layer-wise search-based merging of SAM, MedSAM, and MedicoSAM improves average Dice on 25 tasks by 6.67 points when tuned per task and 4.37 points when tuned across tasks.