AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Perception Prioritized Training of Diffusion Models

Jooyoung Choi, Jungbeom Lee, Chaehun Shin, Sungwon Kim, Hyunwoo Kim, Sungroh Yoon

CVPR 2022 · Image generation › Fast diffusion sampling · approach: Denoising diffusion probabilistic models · 192 citations (OpenAlex)

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Replace baseline loss weighting with P2 weightingADM diffusion models · FFHQ, CUB, AFHQ-Dog, Oxford Flowers, and MetFacesFID-50k: 7.86 -> 6.92; 9.60 -> 6.95; 12.47 -> 11.55; 20.01 -> 17.29; 44.34 -> 36.80Improved
Replace baseline loss weighting with P2 weightingADM diffusion models · FFHQ, CUB, AFHQ-Dog, Oxford Flowers, and MetFacesKID-50k (×103): 3.85 -> 3.46; 3.49 -> 2.38; 4.79 -> 4.10; 16.8 -> 14.8; 22.1 -> 17.6Improved
Use P2 weighting instead of baseline weightingADM diffusion model · CelebA-HQ 256×256FID: 7.22 -> 6.91Improved
Use P2 weighting instead of baseline weightingADM diffusion model · Oxford Flowers 256×256FID: 19.60 -> 17.29Improved
Use P2 weighting instead of baseline weightingADM diffusion model · FFHQ 256×256FID: 8.41 -> 6.97; StyleGAN2: 3.73Mixed
Remove self-attention outside the 8×8 bottleneck (ablation)ADM with P2 weighting · MetFaces, 800k imagesFID-10k: 39.51 -> 43.09Hurt
Replace the BigGAN residual block with the residual block of Ho et al. (ablation)ADM with P2 weighting · MetFaces, 800k imagesFID-10k: 41.93 -> 47.37Hurt
Use two residual blocks (ablation)ADM with P2 weighting · MetFaces, 800k imagesFID-10k: 41.93 -> 42.06Hurt
Set learning rate to 2.5e−5 (ablation)ADM with P2 weighting · MetFaces, 800k imagesFID-10k: 41.93 -> 39.51Improved
Use more sampling steps near t = 0 (130-60-60) rather than 250 uniform steps (ablation)Baseline ADM · FFHQFID-10k: 10.88 -> 10.62Improved
Use 60-130-60 rather than 250 uniform sampling steps (ablation)Baseline ADM · FFHQFID-10k: 10.88 -> 12.23Hurt
Increase the P2 exponent γ above 2 (ablation)P2 weighting · Image generationsample qualityHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy