Research map / Image generation / Autoregressive and latent generation
Coherent Semantic Attention for Image Inpainting
Hongyu Liu, Bin Jiang, Yi Xiao, 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 |
|---|---|---|---|
| replace conventional 3x3 conv with the CSA layer (ablation) | CSA inpainting model · Places2 butte | PSNR: 23.71 -> 24.04 | Improved |
| replace contextual attention layer with the CSA layer (ablation) | CSA inpainting model · Places2 butte | PSNR: 23.74 -> 24.04 | Improved |
| add the consistency loss (ablation) | CSA inpainting model · Places2 butte | PSNR: 23.92 -> 24.04 | Improved |
| add the feature patch discriminator in place of patch-only discriminator (ablation) | CSA inpainting model · Places2 canyon | PSNR: 22.12 -> 22.48 | Improved |
| add the feature patch discriminator in place of the conventional SRFeat feature discriminator (ablation) | CSA inpainting model · Places2 canyon | PSNR: 22.16 -> 22.48 | Improved |
| place CSA layer at 3rd down-sample position instead of 4th (ablation) | CSA inpainting model · Places2 canyon | PSNR: 22.14 -> 22.48 | Improved |
| place CSA layer at 2nd rather than 3rd position (ablation) | CSA inpainting model · Places2 canyon | PSNR: 22.48 -> 22.61 | Mixed |
| use CSA instead of Gated Conv for irregular masks | CSA vs GC · CelebA irregular mask 10-20% | PSNR: 31.67 -> 34.69 | Improved |
| use CSA instead of Shift-net for centering mask | CSA vs SH · CelebA centering hole | PSNR: 26.38 -> 26.54 | Improved |
| use CSA instead of Gated Conv for large irregular masks | CSA vs GC · Paris StreetView irregular mask 40-50% | PSNR: 21.33 -> 23.10 | Improved |
Closest papers by meaning
- Semantic Image Inpainting With Deep Generative Models (CVPR 2017 · similarity 0.91)
- Contextual-based Image Inpainting: Infer, Match, and Translate (ECCV 2018 · similarity 0.91)
- Generative Image Inpainting With Contextual Attention (CVPR 2018 · similarity 0.91)
- Learning Pyramid-Context Encoder Network for High-Quality Image Inpainting (CVPR 2019 · similarity 0.90)
- Guidance and Evaluation: Semantic-Aware Image Inpainting for Mixed Scenes (ECCV 2020 · similarity 0.89)
- Prior Guided GAN Based Semantic Inpainting (CVPR 2020 · similarity 0.89)
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.
