AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Patch Diffusion: Faster and More Data-Efficient Training of Diffusion Models

Zhendong Wang, Yifan Jiang, Huangjie Zheng, Peihao Wang, Pengcheng He, Zhangyang Wang, Weizhu Chen, Mingyuan Zhou

NeurIPS 2023 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 32 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

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
Train with patch-wise score matching and coordinate conditioning rather than full-image-only diffusion trainingEDM-DDPM++ on CelebA-64×64 · CelebA-64×64FID: 1.66 -> 1.77Mixed
Train with patch-wise score matching and coordinate conditioning rather than full-image-only diffusion trainingEDM-DDPM++ on FFHQ-64×64 · FFHQ-64×64FID: 2.60 -> 3.11Hurt
Train with patch-wise score matching and coordinate conditioningEDM-DDPM++ on AFHQv2 small datasets · AFHQv2-Cat, AFHQv2-Dog, AFHQv2-WildFID: 4.60 -> 3.11; 4.94 -> 4.80; 2.59 -> 1.93Improved
Use patch-wise trainingEDM-DDPM++ on CelebA-64×64 · CelebA-64×64Train-Time: ∼ 48h -> ∼ 24hImproved
Use patch-wise trainingLDM-ADM on LSUN-Bedroom-256×256 · LSUN-Bedroom-256×256FID: 4.32 -> 2.75Improved
Use patch-wise trainingLDM-ADM on LSUN-Church-256×256 · LSUN-Church-256×256FID: 4.66 -> 2.66Improved
Use patch-wise trainingLDM-ADM on LSUN-Bedroom-256×256 · LSUN-Bedroom-256×256Train-Time: ∼ 8 Days -> ∼ 4 daysImproved
Use patch-wise trainingEDM-ADM on ImageNet-256×256 · ImageNet-256×256FID (with CFG): 10.56 -> 2.72Improved
Use stochastic rather than progressive patch-size scheduling (ablation)Patch Diffusion · CelebA-64×64FID: 2.05 -> 1.66Improved
Use stochastic rather than progressive patch-size scheduling (ablation)Patch Diffusion · FFHQ-64×64FID: 3.85 -> 3.11Improved
Include full-size images during patch training instead of using only patches (p=0.0) (ablation)Patch Diffusion · CelebA-64×64FIDImproved
Increase the share of full-size images from p=0.0 to p=0.1 (ablation)Patch Diffusion · CelebA-64×64FIDImproved

Later papers that compared with this method

2 later accepted papers report Patch Diffusion as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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