Research map / Image generation / GANs
Progressive Reconstruction of Visual Structure for Image Inpainting
Jingyuan Li, Fengxiang He, Lefei Zhang, Bo Du, Dacheng Tao
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 partial-deconvolution upsampling to P-UNet baseline (ablation) | PRVS · Paris Street View, mask 50%-60% | SSIM: 0.697 -> 0.707 | Improved |
| add partial-deconvolution upsampling to P-UNet baseline (ablation) | PRVS · Paris Street View, mask 50%-60% | PSNR: 22.04 -> 22.19 | Improved |
| add VSR layers on top of partial-deconvolution (ablation) | PRVS · Paris Street View, mask 50%-60% | SSIM: 0.707 -> 0.716 | Improved |
| add VSR layers on top of partial-deconvolution (ablation) | PRVS · Paris Street View, mask 50%-60% | PSNR: 22.19 -> 22.31 | Improved |
| add VSR layers on top of partial-deconvolution (ablation) | PRVS · Paris Street View, mask 50%-60% | MAE: 0.0556 -> 0.0545 | Improved |
| add contextual (pixel) attention to PD+VSR (ablation) | PRVS · Paris Street View, mask 50%-60% | SSIM: 0.716 -> 0.724 | Improved |
| add contextual (pixel) attention to PD+VSR (ablation) | PRVS · Paris Street View, mask 50%-60% | PSNR: 22.31 -> 22.48 | Improved |
| add contextual (pixel) attention to PD+VSR (ablation) | PRVS · Paris Street View, mask 50%-60% | MAE: 0.0545 -> 0.0534 | Improved |
| use PRVS instead of P-UNet at small hole ratio | PRVS · Places2 val, mask 10%-20% | PSNR: 27.67 -> 28.87 | Improved |
| use PRVS instead of Edge-Connect at large hole ratio | PRVS · Places2 val, mask 50%-60% | SSIM: 0.629 -> 0.672 | Improved |
| use PRVS instead of Edge-Connect at mask 10%-20% | PRVS · Places2 val | MAE: 0.0151 -> 0.0125 | Improved |
| use edge from Edge-Connect (two cascaded generators) instead of P-UNet | PRVS · Places2 val, mask 30%-40% | PSNR: 22.52 -> 22.53 | No clear effect |
Closest papers by meaning
- High-Fidelity Image Inpainting with GAN Inversion (ECCV 2022 · similarity 0.89)
- PD-GAN: Probabilistic Diverse GAN for Image Inpainting (CVPR 2021 · similarity 0.88)
- StructureFlow: Image Inpainting via Structure-Aware Appearance Flow (ICCV 2019 · similarity 0.88)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.86)
- In-N-Out: Faithful 3D GAN Inversion with Volumetric Decomposition for Face Editing (CVPR 2024 · similarity 0.86)
- Image Inpainting via Iteratively Decoupled Probabilistic Modeling (ICLR 2024 · similarity 0.86)
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.
