Research map / Image generation / Text-to-image diffusion
Inpaint-Anywhere: Zero-Shot Multi-Identity Inpainting with Efficient Diffusion Transformer
Junsheng Luan, Lei Zhao, Wei Xing
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 |
|---|---|---|---|
| add filters B and C to filter A (ablation) | BoundID dataset · BoundID validation | FID: 16.25 -> 8.521 | Improved |
| add filter C (aesthetic) on top of A and B (ablation) | BoundID dataset · BoundID validation | CLIP-I: 0.741 -> 0.823 | Improved |
| add filter B on top of filter A (fewer images) (ablation) | BoundID dataset · BoundID validation | FID: 9.365 -> 16.25 (CLIP-I 0.616 -> 0.741) | Mixed |
| enlarge bounding box by 50% (ablation) | Inpaint-Anywhere · BoundID validation | FID: 10.04 -> 8.521 | Improved |
| enlarge bounding box from 50% to 75% (ablation) | Inpaint-Anywhere · BoundID validation | DINO / CLIP-T: DINO 0.721 -> 0.703; CLIP-T 0.285 -> 0.261 | Hurt |
| use separate IDA and InpA adapters instead of a unified adapter (ablation) | Inpaint-Anywhere · BoundID validation | FID: 11.57 -> 8.521 | Improved |
| add identity adapter IDA (ablation) | Inpaint-Anywhere · BoundID validation | CLIP-I: 0.616 -> 0.823 | Improved |
| add inpainting adapter InpA (ablation) | Inpaint-Anywhere · BoundID validation | FID: 16.25 -> 8.521 | Improved |
| use IMRoPE position shift for identity tokens (ablation) | Inpaint-Anywhere · BoundID validation | CLIP-I: 0.725 -> 0.823 | Improved |
| add position encoder with masked position-aware attention (ablation) | Inpaint-Anywhere · BoundID validation | identity-to-region correspondence | Improved |
| single-pass multi-identity inpainting vs two-pass baselines | Inpaint-Anywhere · DreamBench | FID: 8.761 -> 7.619 | 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.
- EasyControl — EasyControl: Adding Efficient and Flexible Control for Diffusion Transformer (ICCV 2025) · better on 8 of 8 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
- InfiniteYou: Flexible Photo Recrafting While Preserving Your Identity (ICCV 2025 · similarity 0.87)
- MagicMakeup: A Region-Controllable Diffusion Transformer for High-Fidelity Makeup-Transfer (ECCV 2026 · similarity 0.87)
- Towards Enhanced Image Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistency (CVPR 2025 · similarity 0.86)
- IMFine: 3D Inpainting via Geometry-guided Multi-view Refinement (CVPR 2025 · similarity 0.86)
- Don't Look into the Dark: Latent Codes for Pluralistic Image Inpainting (CVPR 2024 · similarity 0.85)
- DesignEdit: Unify Spatial-Aware Image Editing via Training-free Inpainting with a Multi-Layered Latent Diffusion Framework (AAAI 2025 · similarity 0.85)
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.
