Research map / 3D vision / Face and shape
Geometric Representation Learning for Document Image Rectification
Hao Feng, Wengang Zhou, Jiajun Deng, Yuechen Wang, Houqiang 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 encoder and textline extractor without supervision (ablation) | DocGeoNet baseline · DocUNet Benchmark | LD: 8.11 -> 7.99 | Improved |
| add 3D shape supervision (ablation) | DocGeoNet baseline · DocUNet Benchmark | LD: 7.99 -> 7.83 | Improved |
| add textline supervision (ablation) | DocGeoNet baseline · DocUNet Benchmark | LD: 7.99 -> 7.79 | Improved |
| add both 3D shape and textline supervision (ablation) | DocGeoNet baseline · DocUNet Benchmark | LD: 7.99 -> 7.71 | Improved |
| add both supervisions (ablation) | DocGeoNet baseline · DocUNet Benchmark | CER: 0.2031 -> 0.1821 | Improved |
| add the preprocessing module (ablation) | DocGeoNet · DocUNet Benchmark | LD: 8.61 -> 7.71 | Improved |
| add the preprocessing module (ablation) | DocGeoNet · DocUNet Benchmark | MS-SSIM: 0.4843 -> 0.5040 | Improved |
| use learnable upsampling instead of bilinear (ablation) | DocGeoNet · DocUNet Benchmark | LD: 7.77 -> 7.71 | Improved |
| use learnable upsampling instead of bilinear (ablation) | DocGeoNet · DocUNet Benchmark | MS-SSIM: 0.5062 -> 0.5040 | No clear effect |
| compare DocGeoNet with DocTr | DocTr · DIR300 test | LD: 7.21 -> 6.40 | 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.
- DewarpNet — DewarpNet: Single-Image Document Unwarping With Stacked 3D and 2D Regression Networks (ICCV 2019) · 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
- DocScanner: Robust Document Image Rectification with Progressive Learning (IJCV 2025 · similarity 0.89)
- Can You Read Me Now? Content Aware Rectification using Angle Supervision (ECCV 2020 · similarity 0.88)
- Exploiting Vector Fields for Geometric Rectification of Distorted Document Images (ECCV 2018 · similarity 0.88)
- ForCenNet: Foreground-Centric Network for Document Image Rectification (ICCV 2025 · similarity 0.87)
- DocFormerv2: Local Features for Document Understanding (AAAI 2024 · similarity 0.84)
- DocRes: A Generalist Model Toward Unifying Document Image Restoration Tasks (CVPR 2024 · similarity 0.82)
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.
