The similarity between consecutive denoised images during diffusion sampling correlates with artifact presence, enabling a classifier trained on 680 images to detect flawed outputs with 72.35% accuracy.
Denoising dif- fusion probabilistic models
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
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Similarity Trajectories: Linking Sampling Process to Artifacts in Diffusion-Generated Images
The similarity between consecutive denoised images during diffusion sampling correlates with artifact presence, enabling a classifier trained on 680 images to detect flawed outputs with 72.35% accuracy.