Research map / Image generation / Image restoration
Contextual Residual Aggregation for Ultra High-Resolution 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 CRA to aggregate contextual high-frequency residuals into the hole | Contextual Residual Aggregation · Places2 validation, 1024×1024 | FID: 8.347 -> 4.899 | Helped |
| Use CRA to aggregate contextual high-frequency residuals into the hole | Contextual Residual Aggregation · Places2 validation, 2048×2048 | inference time: 219ms -> 37ms | Helped |
| Apply multi-scale attention transfer using shared attention scores | CRA generator · Places2 inpainting | inpainting quality | Helped |
| Share one set of attention scores across transfer levels | CRA generator · Places2 inpainting | memory and computation efficiency | Helped |
| Use Averaging down-sampling instead of other tested down-sampling methods (ablation) | CRA · HD image inpainting | texture coherence | Helped |
| Use Bilinear or Bicubic up-sampling rather than Nearest Neighbor (ablation) | CRA · HD image inpainting | texture coherence | Helped |
| Replace standard gated convolutions with LWGCsc in both coarse and refine networks (ablation) | CRA generator · CelebA-HQ | visual quality | Hurt |
| Use LWGCsc in the coarse network and LWGCpw in the refine network (ablation) | CRA generator · CelebA-HQ | parameter efficiency | Helped |
| Use LWGCsc in the coarse network and LWGCpw rather than standard gated convolution in the refine network (ablation) | CRA generator · CelebA-HQ | visual quality | No clear effect |
Closest papers by meaning
- High-Resolution Image Inpainting Using Multi-Scale Neural Patch Synthesis (CVPR 2017 · similarity 0.89)
- Learning Pyramid-Context Encoder Network for High-Quality Image Inpainting (CVPR 2019 · similarity 0.87)
- RAD: Region-Aware Diffusion Models for Image Inpainting (CVPR 2025 · similarity 0.85)
- HD-Painter: High-Resolution and Prompt-Faithful Text-Guided Image Inpainting with Diffusion Models (ICLR 2025 · similarity 0.85)
- Paint by Inpaint: Learning to Add Image Objects by Removing Them First (CVPR 2025 · 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.