A teacher-student diffusion framework, Adaptively Distilled ControlNet, uses mask-only student generation with teacher-guided noise alignment and adaptive lesion weighting, improving downstream segmentation on KiTS19 and Polyps.
CAAI AIR2, 9150015 (2021)
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
-
Adaptively Distilled ControlNet: Accelerated Training and Superior Sampling for Medical Image Synthesis
A teacher-student diffusion framework, Adaptively Distilled ControlNet, uses mask-only student generation with teacher-guided noise alignment and adaptive lesion weighting, improving downstream segmentation on KiTS19 and Polyps.