Research map / Image generation / Image restoration
Region Normalization for Image Inpainting
Tao Yu, Zongyu Guo, Xin Jin, Shilin Wu, Zhibo Chen, Weiping Li, Zhizheng Zhang, Sen Liu
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 |
|---|---|---|---|
| replace instance normalization with Region Normalization (RN-B + RN-L) (ablation) | EC image generator backbone · Places2 | PSNR: 24.37 -> 25.10 | Improved |
| use RN-B in encoder instead of IN (ablation) | backbone · Places2 | PSNR: 24.37 -> 24.88 | Improved |
| use RN-B in later layers (res-blocks) instead of RN-L/IN (ablation) | Arch.1 · Places2 | PSNR: 24.88 -> 24.41 | Hurt |
| add RN-L in residual blocks (ablation) | Arch.1 · Places2 | PSNR: 24.88 -> 25.02 | Improved |
| add RN-L in decoder (ablation) | Arch.4 · Places2 | PSNR: 25.02 -> 25.10 | Improved |
| use RN-L in encoder instead of RN-B (ablation) | RN model · Places2 | PSNR: 25.10 -> 24.53 | Hurt |
| use full-spatial normalization (IN/BN) instead of no normalization (ablation) | backbone · Places2 | PSNR: 24.47 -> 24.37 (IN), 24.24 (BN) | Hurt |
| set RN-L threshold to 0.8 instead of 0.9 (ablation) | RN · CelebA | PSNR: 23.73 -> 28.51 | Improved |
| apply RN to CA | CA · Places2 | PSNR: 21.60 -> 24.12 | Improved |
| apply RN to PC | PC · Places2 | PSNR: 24.82 -> 25.32 | Improved |
| apply RN to GC | GC · Places2 | PSNR: 24.53 -> 24.55 | No clear effect |
| apply RN with larger mask area | backbone · CelebA | l1(%): -0.03 (0-10%) to -0.31 (50-60%) | Improved |
Closest papers by meaning
- VCNet: A Robust Approach to Blind Image Inpainting (ECCV 2020 · similarity 0.88)
- Image Inpainting With Learnable Bidirectional Attention Maps (ICCV 2019 · similarity 0.87)
- Localization of Deep Inpainting Using High-Pass Fully Convolutional Network (ICCV 2019 · similarity 0.87)
- InpDiffusion: Image Inpainting Localization via Conditional Diffusion Models (AAAI 2025 · similarity 0.85)
- IMFine: 3D Inpainting via Geometry-guided Multi-view Refinement (CVPR 2025 · similarity 0.84)
- Towards Enhanced Image Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistency (CVPR 2025 · similarity 0.84)
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.
