Research map / Image generation / Editing and style transfer
UNet-Twice: A Simple Structured Reference-based Inpainting Framework
Junda Lu, zhiqiao xu, Houkun Wu, Wei Cui, Bo Huang, Mingyang Chen, Bing Li
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 |
|---|---|---|---|
| inject only 1 deep feature (n=1) instead of 13 from UNet-I (ablation) | UNet-Twice · MegaDepth (1 reference) | PSNR: 26.9106 -> 27.3858 | Improved |
| inject only 1 deep feature (n=1) instead of 2 (ablation) | UNet-Twice · MegaDepth (1 reference) | DreamSim: 0.0166 -> 0.0106 | Improved |
| use fine or simple homography variants instead of base (ablation) | UNet-Twice · MegaDepth (1 reference) | PSNR: 27.3855 -> 27.4491 (simple), 27.4300 (fine) | No clear effect |
| use affine instead of homography alignment (ablation) | UNet-Twice · MegaDepth (1 reference) | PSNR: 27.3855 -> 27.2807 | No clear effect |
| use UNet-Twice instead of LeftRefill | UNet-Twice · MegaDepth (1 reference) | PSNR: 26.0812 -> 27.3856 | Improved |
| increase references from 1 to 4 | UNet-Twice · MegaDepth | PSNR: 27.3856 -> 28.5596 | Improved |
| increase references from 2 to 4 | UNet-Twice · MegaDepth | PSNR: 28.4582 -> 28.5596 | No clear effect |
| increase references from 2 to 4 | RealFill · MegaDepth | PSNR: 26.2415 -> 26.6568 | No clear effect |
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.
- LeftRefill — LeftRefill: Filling Right Canvas based on Left Reference through Generalized Text-to-Image Diffusion Model (CVPR 2024) · better on 7 of 7 shared results
- CompleteMe — CompleteMe: Reference-based Human Image Completion (ICCV 2025) · better on 2 of 2 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
- Towards Enhanced Image Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistency (CVPR 2025 · similarity 0.85)
- Continuously Masked Transformer for Image Inpainting (ICCV 2023 · similarity 0.85)
- IMFine: 3D Inpainting via Geometry-guided Multi-view Refinement (CVPR 2025 · similarity 0.84)
- Shadow-Enlightened Image Outpainting (CVPR 2024 · similarity 0.84)
- AuraFusion360: Augmented Unseen Region Alignment for Reference-based 360deg Unbounded Scene Inpainting (CVPR 2025 · similarity 0.83)
- Keyframe-Guided Creative Video Inpainting (CVPR 2025 · similarity 0.83)
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.
