Research map / Image generation / Editing and style transfer
OSOR: One-Step Diffusion Inpainting for Effect-Aware Object Removal
Qinming Zhou, Chenxi Sun, Deyang Kong, Junhao He, Xiangheng Tang, Peike Yu, Haotian Wu, Leilei Cao, Linfeng Zhang
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 noise level t=400 instead of t=200 for one-step restoration (ablation) | OSOR(SDXL) · RORD-Val (object-effect mask) | FID: 60.1520 -> 50.0639 | Improved |
| use noise level t=400 instead of t=600 (ablation) | OSOR(SDXL) · RORD-Val (object-effect mask) | FID: 55.4021 -> 50.0639 | Improved |
| use noise level t=400 instead of t=800 (ablation) | OSOR(SDXL) · RORD-Val (object-effect mask) | PSNR: 22.644 -> 22.696 | Improved |
| occupancy (area-pooled) patch targets instead of hard mask targets (ablation) | OSOR(SDXL) · RORD-Val (object-effect mask) | FID: 56.1508 -> 50.0639 | Improved |
| occupancy patch targets instead of Gaussian soft targets (ablation) | OSOR(SDXL) · RORD-Val (object-effect mask) | FID: 64.1290 -> 50.0639 | Improved |
| occupancy patch targets instead of Gaussian soft targets (ablation) | OSOR(SDXL) · RORD-Val (object-effect mask) | PSNR: 21.992 -> 22.696 | Improved |
| Gaussian soft targets instead of hard mask targets (ablation) | OSOR(SDXL) · RORD-Val (object-effect mask) | FID: 56.1508 -> 64.1290 | Hurt |
| add alpha head with alpha compositing (Phase II) over hard blending (Phase I) (ablation) | OSOR · RORD-Val (object-only and effect-aware conditioning masks) | fidelity and perceptual metrics | Improved |
| real-pair refinement on captured OBER subset for 1K steps | OSOR(FLUX) · RORD-Val | FID: 27.4/28.1 -> 23.8/23.8 | Improved |
| real-pair refinement on captured OBER subset | OSOR(FLUX) · OmniPaint-Bench | FID: 49.2 -> 44.4 | Improved |
| real-pair refinement on captured OBER subset | OSOR(FLUX) · RemovalBench | FID: 43.9 -> 42.5 | Improved |
| switch conditioning mask from object-only to object-effect mask | OSOR(FLUX) · CORNE-Val | FID: 12.3241 -> 12.5203 | No clear effect |
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.
- OmniPaint — OmniPaint: Mastering Object-Oriented Editing via Disentangled Insertion-Removal Inpainting (ICCV 2025) · better on 8 of 9 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
- SmartEraser: Remove Anything from Images using Masked-Region Guidance (CVPR 2025 · similarity 0.89)
- In-N-Out: Lifting 2D Diffusion Prior for 3D Object Removal via Tuning-Free Latents Alignment (NeurIPS 2024 · similarity 0.89)
- Paint by Inpaint: Learning to Add Image Objects by Removing Them First (CVPR 2025 · similarity 0.88)
- Attentive Eraser: Unleashing Diffusion Model’s Object Removal Potential via Self-Attention Redirection Guidance (AAAI 2025 · similarity 0.87)
- Instruct2See: Learning to Remove Any Obstructions Across Distributions (ICML 2025 · similarity 0.87)
- RORem: Training a Robust Object Remover with Human-in-the-Loop (CVPR 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.
