Research map / Image generation / Image restoration
Rethinking Image Inpainting via a Mutual Encoder-Decoder with Feature Equalizations
Hongyu Liu, Bin Jiang, Yibing Song, Wei Huang, Chao Yang
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 texture branch (ablation) | MEDFE · Paris StreetView | FID: 30.37 -> 25.10 | Improved |
| add structure branch (ablation) | MEDFE · Paris StreetView | FID: 27.46 -> 25.10 | Improved |
| add texture branch (ablation) | MEDFE · Paris StreetView | PSNR: 22.80 -> 23.38 | Improved |
| add structure branch (ablation) | MEDFE · Paris StreetView | PSNR: 22.96 -> 23.38 | Improved |
| add feature equalization (ablation) | MEDFE · Place2 | FID: 29.11 -> 21.26 | Improved |
| add feature equalization (ablation) | MEDFE · Place2 | PSNR: 23.14 -> 24.57 | Improved |
| add feature equalization (ablation) | MEDFE · Place2 | SSIM: 0.837 -> 0.852 | Improved |
| use bilateral propagation equalization instead of non-local aggregation (ablation) | MEDFE · Place2 | FID: 24.07 -> 21.26 | Improved |
| use non-local aggregation instead of no equalization (ablation) | MEDFE · Place2 | FID: 29.11 -> 24.07 | Improved |
| use MEDFE instead of CSA at 10-20% masks | MEDFE · Place2 (10-20% mask) | PSNR: 31.31 -> 31.13 | No clear effect |
| use MEDFE instead of CSA at 40-50% masks | MEDFE · Place2 (40-50% mask) | FID: 38.74 -> 28.79 | Improved |
| use MEDFE instead of SH on center holes | MEDFE · CelebA | FID: 29.72 -> 25.51 | Improved |
Closest papers by meaning
- StructureFlow: Image Inpainting via Structure-Aware Appearance Flow (ICCV 2019 · similarity 0.88)
- Parallel Multi-Resolution Fusion Network for Image Inpainting (ICCV 2021 · similarity 0.88)
- Generative Image Inpainting With Contextual Attention (CVPR 2018 · similarity 0.88)
- Image Inpainting With Learnable Bidirectional Attention Maps (ICCV 2019 · similarity 0.88)
- Learning Pyramid-Context Encoder Network for High-Quality Image Inpainting (CVPR 2019 · similarity 0.88)
- SyFormer: Structure-Guided Synergism Transformer for Large-Portion Image Inpainting (AAAI 2024 · 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.
