AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

SnapFusion: Text-to-Image Diffusion Model on Mobile Devices within Two Seconds

Yanyu Li, Huan Wang, Qing Jin, Ju Hu, Pavlo Chemerys, Yun Fu, Yanzhi Wang, Sergey Tulyakov, Jian Ren

NeurIPS 2023 · Image generation › Text-to-image diffusion · approach: Text-driven synthesis and semantic guidance · 35 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
use CFG-aware step distillation loss alongside vanilla distillation (ablation)8-step efficient UNet · MS-COCO 2014 valCLIP score: +0.002 to +0.007 CLIP over SD-v1.5 at similar FIDImproved
vanilla step distillation only (no CFG-aware loss) (ablation)8-step student · MS-COCO 2014 valCLIP score: capped at 0.30Hurt
mix vanilla and CFG distillation losses (ablation)step distillation · MS-COCO 2014 val 6KFID and CLIPImproved
direct distillation from 32-step teacher instead of progressive distillation (ablation)step distillation of SD-v1.5 · MS-COCO 2014 val 6KFID and CLIPImproved
add original denoising loss during CFG distillation (ablation)CFG distillation · MS-COCO 2014 val 6KFIDImproved
use more teacher steps (32/64 instead of 16) for 8-step student (ablation)vanilla step distillation · MS-COCO 2014 val 6KCLIP scoreHurt
increase CFG probability and CFG range (ablation)CFG distillation · MS-COCO 2014 val 6KFID vs CLIPMixed
add robust training (stochastic block skipping) before evolving architecture (ablation)SD-v1.5 UNet · qualitative block removalimage quality after removing blocksImproved
distill to 8 steps with CFG-aware loss vs w-conditioned distillation at 16 steps (ablation)SD-v1.5 distillation · MS-COCO 2014 val 6KFID and CLIPImproved
reduce UNet steps from 16 to 8 without distillation (ablation)efficient UNet · MS-COCO 2014 val 6KFID and CLIPHurt
apply CFG-aware distillation to SD-v2 for 8 stepsSD-v2 · MS-COCO 2014 valFID and CLIPNo clear effect

Later papers that compared with this method

2 later accepted papers report SnapFusion 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