Research map / Architectures and efficiency / CNN design
Recurrent Feature Reasoning for Image Inpainting
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 recurrent feature reasoning with KCA | RFR-Net · Places2, CelebA, and Paris Street View, 50%-60% masks | SSIM: 0.574 -> 0.596 (Places2); 0.799 -> 0.819 (CelebA); 0.659 -> 0.681 (Paris Street View) | Helped |
| Use recurrent feature reasoning with KCA | RFR-Net · Places2, CelebA, and Paris Street View, 50%-60% masks | PSNR: 18.67 -> 18.92 (Places2); 22.37 -> 22.88 (CelebA); 22.07 -> 22.40 (Paris Street View) | Helped |
| Remove the attention module (ablation) | RFR-Net · Paris Street View, 40%-50% mask | SSIM | Hurt |
| Increase recurrence count from 6 to 7 (ablation) | RFR-Net · Paris Street View | SSIM: 0.857 -> 0.852 | Hurt |
| Increase recurrence count from 6 to 8 (ablation) | RFR-Net · Paris Street View | SSIM: 0.857 -> 0.854 | Hurt |
| Move RFR deeper, after downsampling twice rather than once (ablation) | RFR-Net · Paris StreetView, 30%-40% mask | SSIM: 0.860 -> 0.845 | Hurt |
| Move RFR deeper, after downsampling three times rather than once (ablation) | RFR-Net · Paris StreetView, 30%-40% mask | SSIM: 0.860 -> 0.839 | Hurt |
| Use adaptive feature merging instead of average merging (ablation) | RFR-Net · Paris StreetView | visual detail quality | Helped |
| Use intermediate feature maps for adaptive merging instead of only the last feature map (ablation) | RFR-Net · Paris StreetView | visual texture quality | Helped |
| Replace Edge-Connect's structure generator residual blocks with the RFR module (ablation) | Edge-Connect structure generator · Structure estimation | visual structure quality | Helped |
Closest papers by meaning
- MAT: Mask-Aware Transformer for Large Hole Image Inpainting (CVPR 2022 · similarity 0.87)
- Foreground-Aware Image Inpainting (CVPR 2019 · similarity 0.86)
- Reconstructing the Image Stitching Pipeline: Integrating Fusion and Rectangling into a Unified Inpainting Model (NeurIPS 2024 · similarity 0.86)
- Learning to Incorporate Structure Knowledge for Image Inpainting (AAAI 2020 · similarity 0.86)
- HD-Painter: High-Resolution and Prompt-Faithful Text-Guided Image Inpainting with Diffusion Models (ICLR 2025 · similarity 0.85)
- SyFormer: Structure-Guided Synergism Transformer for Large-Portion Image Inpainting (AAAI 2024 · 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.