Research map / Image generation / Editing and style transfer
OmniPaint: Mastering Object-Oriented Editing via Disentangled Insertion-Removal Inpainting
Yongsheng Yu, Ziyun Zeng, Haitian Zheng, Jiebo Luo
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use OmniPaint instead of FreeCompose for removal | object removal · Object removal (300-sample real-world test) | FID: 88.77 -> 51.66 | Improved |
| use OmniPaint instead of FreeCompose for removal | object removal · Object removal (300-sample real-world test) | CFD: 0.3743 -> 0.2619 | Improved |
| use OmniPaint instead of FreeCompose for removal | object removal · Object removal (300-sample real-world test) | ReMOVE: 0.8654 -> 0.8610 | No clear effect |
| use OmniPaint instead of PowerPaint for removal | object removal · Object removal (RORD 1000 samples) | FID: 42.65 -> 19.17 | Improved |
| use OmniPaint instead of PowerPaint for removal | object removal · Object removal (RORD 1000 samples) | LPIPS: 0.0968 -> 0.0424 | Improved |
| use OmniPaint instead of IMPRINT for insertion | object insertion · Object insertion (565 samples) | DINOv2: 76.8940 -> 84.3738 | Improved |
| use OmniPaint instead of IMPRINT for insertion | object insertion · Object insertion (565 samples) | DreamSim: 0.1854 -> 0.1557 | Improved |
| use OmniPaint instead of PbE for image quality | object insertion · Object insertion (565 samples) | MUSIQ: 70.26 -> 70.59 | Improved |
| increase cycle loss weight gamma from 0 to 1.5 (ablation) | OmniPaint insertion · Cycle loss weight (qualitative) | physical effect synthesis | Improved |
| increase cycle loss weight gamma from 1.5 to 3.0 (ablation) | OmniPaint insertion · Cycle loss weight (qualitative) | physical effect realism | Hurt |
| increase NFE from 18 to 28 (ablation) | OmniPaint · NFE (qualitative) | visual quality | No clear effect |
| increase NFE from 1 or 4 to 18 (ablation) | OmniPaint · NFE (qualitative) | blurring in masked regions | Improved |
Later papers that compared with this method
6 later accepted papers report OmniPaint as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- OSOR: One-Step Diffusion Inpainting for Effect-Aware Object Removal (ECCV 2026 · better on 8 of 9 shared results)
- Precise Object and Effect Removal with Adaptive Target-Aware Attention (CVPR 2026 · better on 5 of 6 shared results)
- RevealLayer: Disentangling Hidden and Visible Layers via Occlusion-Aware Image Decomposition (ICML 2026 · better on 4 of 5 shared results)
- PICS: Pairwise Image Compositing with Spatial Interactions (ICLR 2026 · better on 3 of 5 shared results)
- You Only Erase Once: Erasing Anything without Bringing Unexpected Content (CVPR 2026 · better on 5 of 5 shared results)
- PixPerfect: Seamless Latent Diffusion Local Editing with Discriminative Pixel-Space Refinement (NeurIPS 2025 · better on 1 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FreeCompose — FreeCompose: Generic Zero-Shot Image Composition with Diffusion Prior (ECCV 2024) · better on 6 of 7 shared results
- IMPRINT — IMPRINT: Generative Object Compositing by Learning Identity-Preserving Representation (CVPR 2024) · better on 4 of 4 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
- DesignEdit: Unify Spatial-Aware Image Editing via Training-free Inpainting with a Multi-Layered Latent Diffusion Framework (AAAI 2025 · similarity 0.89)
- Paint by Inpaint: Learning to Add Image Objects by Removing Them First (CVPR 2025 · similarity 0.89)
- Generative Omnimatte: Learning to Decompose Video into Layers (CVPR 2025 · similarity 0.88)
- Towards Enhanced Image Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistency (CVPR 2025 · similarity 0.88)
- OmniEdit: Building Image Editing Generalist Models Through Specialist Supervision (ICLR 2025 · similarity 0.87)
- A Diffusion-Based Framework for Occluded Object Movement (AAAI 2025 · similarity 0.87)
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.
