The paper reports a training-free extrapolation method that combines per-patch prompt selection from cross-attention masks, ControlNet edge guidance, and permuted dilated sampling to reduce repetition and distortion at 4x to 16x upscales.
Stable diffusion v2-1 model card,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
AccDiffusion v2: Towards More Accurate Higher-Resolution Diffusion Extrapolation
The paper reports a training-free extrapolation method that combines per-patch prompt selection from cross-attention masks, ControlNet edge guidance, and permuted dilated sampling to reduce repetition and distortion at 4x to 16x upscales.