Research map / Image generation / Autoregressive and latent generation
Cross-Image Context for Single Image Inpainting
Tingliang Feng, Wei Feng, Weiqi Li, Di Lin
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 |
|---|---|---|---|
| remove intra-set similarity (ablation) | CICM · Places2 mask 40-60% | PSNR: 19.211 -> 16.647 | Hurt |
| remove inter-set similarity (ablation) | CICM · Places2 mask 40-60% | PSNR: 19.211 -> 16.338 | Hurt |
| update only top-1 cross-image feature instead of all in relevant set (ablation) | CICM · Places2 mask 40-60% | PSNR: 19.211 -> 16.632 | Hurt |
| update only top-50% cross-image features (ablation) | CICM · Places2 mask 40-60% | PSNR: 19.211 -> 17.118 | Hurt |
| add corrupted-ratio gating of context propagation (ablation) | CICM · Places2 mask 40-60% | PSNR: 18.879 -> 19.211 | Improved |
| use ground-truth corrupted ratios instead of estimated (ablation) | CICM · Places2 mask 40-60% | PSNR: 19.211 -> 19.262 | No clear effect |
| augment with all cross-image features rather than top-1 (ablation) | CICM · Places2 mask 40-60% | PSNR: 17.866 -> 19.211 | Improved |
| augment with all cross-image features rather than top-50% (ablation) | CICM · Places2 mask 40-60% | PSNR: 18.295 -> 19.211 | Improved |
| cross-image context instead of single-image clustering (k-means, RUC, anchor only) (ablation) | CICM · Places2 mask 40-60% | PSNR: 17.235 -> 19.211 | Improved |
| use anchor features to select context rather than merged sets (ablation) | CICM · Places2 mask 40-60% | PSNR: 17.574 -> 19.211 | Improved |
| add CICM module to MISF | MISF · Places2 mask 40-60% | PSNR: 19.314 -> 21.267 | Improved |
| add CICM module to RFR | RFR · CelebA mask 40-60% | PSNR: 20.133 -> 21.517 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MISF — MISF: Multi-Level Interactive Siamese Filtering for High-Fidelity Image Inpainting (CVPR 2022) · better on 8 of 8 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
- Contextual-based Image Inpainting: Infer, Match, and Translate (ECCV 2018 · similarity 0.89)
- Learning Pyramid-Context Encoder Network for High-Quality Image Inpainting (CVPR 2019 · similarity 0.87)
- Context Encoders: Feature Learning by Inpainting (CVPR 2016 · similarity 0.87)
- CR-Fill: Generative Image Inpainting With Auxiliary Contextual Reconstruction (ICCV 2021 · similarity 0.87)
- Common Inpainted Objects In-N-Out of Context (CVPR 2026 · similarity 0.85)
- Brush2Prompt: Contextual Prompt Generator for Object Inpainting (CVPR 2024 · 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.
