Refine-by-Align uses diffusion cross-attention maps to locate the reference region matching a masked artifact, then re-inpaints the artifact with that reference detail.
Multi-concept customization of text-to-image diffusion
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
-
Refine-by-Align: Reference-Guided Artifacts Refinement through Semantic Alignment
Refine-by-Align uses diffusion cross-attention maps to locate the reference region matching a masked artifact, then re-inpaints the artifact with that reference detail.