A single diffusion L2 loss can train scalable image tokenizers that match or outperform GAN-LPIPS tokenizers for reconstruction and downstream generation.
Pros and cons of gan evaluation measures
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
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Diffusion Autoencoders are Scalable Image Tokenizers
A single diffusion L2 loss can train scalable image tokenizers that match or outperform GAN-LPIPS tokenizers for reconstruction and downstream generation.