Research map / 3D vision / Face and shape
DewarpNet: Single-Image Document Unwarping With Stacked 3D and 2D Regression Networks
Sagnik Das, Ke Ma, Zhixin Shu, Dimitris Samaras, Roy Shilkrot
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 CoordConv to texture mapping network (ablation) | DewarpNet · Doc3D validation | l2 error on B: 4.73 × 10−5 -> 3.99 × 10−5 | Improved |
| add CoordConv to texture mapping network (ablation) | DewarpNet · Doc3D validation | SSIM on D: 0.9260 -> 0.9281 | Improved |
| add LD loss to LB (ablation) | DewarpNet · Doc3D validation | l2 error on B: 1.40 × 10−4 -> 3.99 × 10−5 | Improved |
| add LD loss to LB (ablation) | DewarpNet · Doc3D validation | SSIM on D: 0.8539 -> 0.9281 | Improved |
| train DocUNet on Doc3D instead of its 2D synthetic dataset | DocUNet · DocUNet benchmark | Local Distortion: 14.08 -> 10.85 | Improved |
| train DocUNet on Doc3D instead of its 2D synthetic dataset | DocUNet · DocUNet benchmark | MS-SSIM: 0.41 -> 0.4389 | Improved |
| use DewarpNet (3D shape then texture mapping) instead of DocUNet on Doc3D | DewarpNet · DocUNet benchmark | Local Distortion: 10.90 -> 8.98 | Improved |
| add refinement (shading removal) network to DewarpNet | DewarpNet · DocUNet benchmark | MS-SSIM: 0.4692 -> 0.4735 | Improved |
| add refinement (shading removal) network to DewarpNet | DewarpNet · DocUNet benchmark | Local Distortion: 8.98 -> 8.95 | No clear effect |
| add refinement network to DewarpNet | DewarpNet · OCR (25 images) | CER: 0.3136 -> 0.2692 | Improved |
| use DewarpNet instead of DocUNet | DewarpNet · OCR (25 images) | CER: 0.4656 -> 0.3136 | Improved |
| use DewarpNet on the most challenging class (f) instead of DocUNet | DewarpNet · DocUNet benchmark class (f) | MS-SSIM and LD | No clear effect |
Later papers that compared with this method
2 later accepted papers report DewarpNet as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Learning an Isometric Surface Parameterization for Texture Unwrapping (ECCV 2022 · better on 11 of 11 shared results)
- Geometric Representation Learning for Document Image Rectification (ECCV 2022 · 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
- D2Dewarp: Dual Dimensions Geometric Representation Learning Based Document Image Dewarping (CVPR 2026 · similarity 0.88)
- DocUNet: Document Image Unwarping via a Stacked U-Net (CVPR 2018 · similarity 0.88)
- Axis-Aligned Document Dewarping (AAAI 2026 · similarity 0.87)
- Revisiting Document Image Dewarping by Grid Regularization (CVPR 2022 · similarity 0.86)
- DocFormerv2: Local Features for Document Understanding (AAAI 2024 · similarity 0.81)
- DocRes: A Generalist Model Toward Unifying Document Image Restoration Tasks (CVPR 2024 · similarity 0.81)
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.
