PatchICL reduces compute by 44% in in-context medical image segmentation at 512x512 resolution while matching a strong global-attention baseline on in-domain CT and outperforming it on 6 of 13 out-of-domain modality categories.
Show and segment: Universal medical image segmentation via in-context learning.arXiv preprint
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Scaling In-Context Segmentation with Hierarchical Supervision
PatchICL reduces compute by 44% in in-context medical image segmentation at 512x512 resolution while matching a strong global-attention baseline on in-domain CT and outperforming it on 6 of 13 out-of-domain modality categories.