Research map / Image generation / Editing and style transfer
StructureFlow: Image Inpainting via Structure-Aware Appearance Flow
Yurui Ren, Xiaoming Yu, Ruonan Zhang, Thomas H. Li, Shan Liu, Ge Li
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 structure reconstructor stage (ablation) | StructureFlow · Paris | PSNR: 28.46 -> 29.25 | Improved |
| add structure reconstructor stage (ablation) | StructureFlow · Celeba | PSNR: 29.42 -> 30.31 | Improved |
| train structure reconstructor on edge-preserved smooth images instead of ground truth images (ablation) | StructureFlow · Paris | PSNR: 28.41 -> 29.25 | Improved |
| train structure reconstructor on edge-preserved smooth images instead of ground truth images (ablation) | StructureFlow · Celeba | PSNR: 29.61 -> 30.31 | Improved |
| add appearance flow blocks in texture generator (ablation) | StructureFlow · Paris | PSNR: 28.77 -> 29.25 | Improved |
| add appearance flow blocks in texture generator (ablation) | StructureFlow · Celeba | PSNR: 29.91 -> 30.31 | Improved |
| add appearance flow blocks in texture generator (ablation) | StructureFlow · Paris | SSIM: 0.8906 -> 0.8979 | Improved |
| smoothing strength sigma 0 to 3 (ablation) | StructureFlow · Paris | PSNR: 28.41 -> 29.25 | Improved |
| increase smoothing strength sigma 3 to 9 (ablation) | StructureFlow · Paris | PSNR: 29.25 -> 28.98 | Hurt |
| use Gaussian sampling instead of bilinear sampling in appearance flow (ablation) | StructureFlow · Places2 flow fields | flow quality | Improved |
| add sampling correctness loss (ablation) | StructureFlow · Places2 flow fields | flow quality | Improved |
| StructureFlow versus EdgeConnect on Places2 at mask 40-60% | StructureFlow · Places2 (mask 40-60%) | FID: 22.3803 vs 19.0003 | Mixed |
Later papers that compared with this method
2 later accepted papers report StructureFlow as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- ShiftDDPMs: Exploring Conditional Diffusion Models by Shifting Diffusion Trajectories (AAAI 2023 · better on 3 of 3 shared results)
- Bridging the Visual Gap: Wide-Range Image Blending (CVPR 2021 · better on 2 of 2 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 0.90)
- Generative Image Inpainting With Contextual Attention (CVPR 2018 · similarity 0.90)
- Learning to Incorporate Structure Knowledge for Image Inpainting (AAAI 2020 · similarity 0.90)
- Image Inpainting via Conditional Texture and Structure Dual Generation (ICCV 2021 · similarity 0.90)
- An Internal Learning Approach to Video Inpainting (ICCV 2019 · similarity 0.89)
- Incremental Transformer Structure Enhanced Image Inpainting With Masking Positional Encoding (CVPR 2022 · 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.
