A training-free colorization method repurposing Stable Diffusion self-attention as cross-image attention, with dual attention and classifier-free guidance, reports the best FID and SI-FID on standard benchmarks.
Title resolution pending
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
-
Leveraging the Powerful Attention of a Pre-trained Diffusion Model for Exemplar-based Image Colorization
A training-free colorization method repurposing Stable Diffusion self-attention as cross-image attention, with dual attention and classifier-free guidance, reports the best FID and SI-FID on standard benchmarks.