DynBC evaluates candidate model updates by their agreement with the current model on augmented reference patches, filtering out updates that deviate too much, and reports improved robustness to client drift and catastrophic forgetting in histopathology segmentation.
Albumentations: fast and flexible image augmenta- tions
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Federated-Continual Dynamic Segmentation of Histopathology guided by Barlow Continuity
DynBC evaluates candidate model updates by their agreement with the current model on augmented reference patches, filtering out updates that deviate too much, and reports improved robustness to client drift and catastrophic forgetting in histopathology segmentation.