LAFR uses a 1024-entry codebook adapter to map low-quality face latents into the high-quality latent space of Stable Diffusion, then LoRA-tunes a pruned UNet on just 600 FFHQ images for blind face restoration.
Accurate 3d face reconstruc- tion with weakly-supervised learning: From single image to image set
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
REJECT 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
LAFR: Efficient Diffusion-based Blind Face Restoration via Latent Codebook Alignment Adapter
LAFR uses a 1024-entry codebook adapter to map low-quality face latents into the high-quality latent space of Stable Diffusion, then LoRA-tunes a pruned UNet on just 600 FFHQ images for blind face restoration.