ReGA adapts a source-trained pelvic bone segmentation model to unseen CT domains by gating pseudo-labels, feature refinement, and teacher updates on a dropout-based reliability score.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Test-time Adaptation of Pelvic Bone Segmentation Models via Dynamic Reliability-Guided
ReGA adapts a source-trained pelvic bone segmentation model to unseen CT domains by gating pseudo-labels, feature refinement, and teacher updates on a dropout-based reliability score.