AI Paper Atlas

Research map / Image generation / Editing and style transfer

TurboEdit: Real-time text-based disentangled real image editing

Zongze Wu, Nicholas I Kolkin, Jonathan Brandt, Richard Zhang, Eli Shechtman

ECCV 2024 · Image generation › Editing and style transfer · approach: Inversion-based diffusion editing

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 encoder-based iterative inversion on SDXL-Turbo instead of multi-step diffusion editing baselinesTurboEdit · PIE-Bench (descriptive prompts)PSNR: 28.51 -> 29.52 vs Inversion-FreeImproved
use 4 inversion steps instead of DDIM at 4 stepsfew-step inversion · PIE-Bench (descriptive prompts)PSNR: 18.59 -> 29.52Improved
increase the number of inversion steps from single-step to multi-step (ablation)TurboEdit inversion network · 10k validation imagesreconstruction qualityImproved
condition on a detailed text prompt instead of a short one (ablation)TurboEdit · qualitativestructure preservation and disentanglementImproved
add rough attention-derived local mask (ablation)TurboEdit · qualitativebackground and identity preservationImproved
freeze attention maps at all steps in few-step diffusion models (ablation)TurboEdit · qualitativestructure and artifactsHurt
add KL regularization on predicted noise (lambda 1e-6)inversion network · qualitativeartifacts and editabilityImproved

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