AI Paper Atlas

Research map / Image generation / Editing and style transfer

ObjectDrop: Bootstrapping Counterfactuals for Photorealistic Object Removal and Insertion

Daniel Winter, Matan Cohen, Shlomi Fruchter, Yael Pritch, Alex Rav-Acha, Yedid Hoshen

ECCV 2024 · Image generation › Editing and style transfer · approach: Inversion-based diffusion editing · 57 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
Fine-tune on counterfactual before/after pairs instead of using a pre-trained inpainting modelInpainting diffusion model · Object removal, held-out test setPSNR: 21.192 -> 23.153Improved
Add bootstrap supervision (pre-train on 350K synthetic counterfactuals) before fine-tuning on real pairs (ablation)ObjectDrop insertion model · Object insertion, held-out test setPSNR: 20.178 -> 21.625Improved
Add bootstrap supervision (ablation)ObjectDrop insertion model · Object insertion, held-out test setLPIPS: 0.066 -> 0.057Improved
Increase counterfactual dataset size for removal (ablation)ObjectDrop removal model · Object removalvisual qualityImproved
Pre-train from text-to-image model instead of inpainting model (ablation)ObjectDrop removal · Object removal, shadows and reflectionsvisual qualityNo clear effect
Apply counterfactual fine-tuning on public Stable-Diffusion-Inpainting v1SD Inpainting v1 · Object removal, held-out test setPSNR: 19.198 -> 21.363Improved
Train insertion on the 2,500-pair real dataset only (ablation)ObjectDrop insertion · Object insertionvisual qualityHurt
Use ObjectDrop instead of Emu Edit and MGIE for removalObjectDrop · Emu Edit benchmark user studyuser preference: 35.9% -> 64.1% (Emu Edit); 13.5% -> 86.5% (MGIE)Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

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