Research map / Image generation / GANs
DE-GAN: A Conditional Generative Adversarial Network for Document Enhancement
Mohamed Ali Souibgui, Yousri Kessentini
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 adversarial training (DE-GAN) to the U-net generator (ablation) | U-net document cleaner · Noisy Office | PSNR: 36.02 -> 38.12 | Improved |
| add adversarial training (DE-GAN) to the U-net generator (ablation) | U-net document cleaner · Noisy Office | SSIM: 0.9970 -> 0.9986 | Improved |
| use DE-GAN instead of the strongest prior method Vo et al. | DE-GAN · DIBCO 2013 | F-measure: 94.4 -> 99.5 | Improved |
| use DE-GAN instead of the DIBCO 2017 winner | DE-GAN · DIBCO 2017 | F-measure: 91.04 -> 97.91 | Improved |
| use DE-GAN instead of the H-DIBCO 2018 winner, which uses pre- and post-processing | DE-GAN · H-DIBCO 2018 | F-measure: 88.34 -> 77.59 | Hurt |
| use DE-GAN instead of the best prior watermark removal method (Dekel et al.) | DE-GAN · Watermark removal | PSNR: 36.02 -> 40.98 | Improved |
| use DE-GAN instead of cycleGAN | DE-GAN · H-DIBCO 2018 | F-measure: 56.33 -> 77.59 | Improved |
| use DE-GAN instead of pix2pix-HD | DE-GAN · H-DIBCO 2018 | F-measure: 72.79 -> 77.59 | Improved |
| use DE-GAN instead of pix2pix-HD on small 300x300 patches | DE-GAN · Document deblurring | PSNR: 19.89 -> 20.37 | Improved |
| use GAN models instead of a plain CNN for deblurring | deblurring · Document deblurring | PSNR: 19.36 -> 20.37 | Improved |
| apply DE-GAN enhancement before Tesseract OCR | Tesseract OCR · 4 degraded/watermarked images | character error rate: 0.37 -> 0.01 | Improved |
Closest papers by meaning
- Single Image Dehazing via Conditional Generative Adversarial Network (CVPR 2018 · similarity 0.87)
- Learning to Super-Resolve Blurry Face and Text Images (ICCV 2017 · similarity 0.85)
- DocNLC: A Document Image Enhancement Framework with Normalized and Latent Contrastive Representation for Multiple Degradations (AAAI 2024 · similarity 0.85)
- Physics-Based Generative Adversarial Models for Image Restoration and Beyond (TPAMI 2020 · similarity 0.85)
- E$^2$GAN: Efficient Training of Efficient GANs for Image-to-Image Translation (ICML 2024 · similarity 0.83)
- Dehaze-RetinexGAN: Real-World Image Dehazing via Retinex-based Generative Adversarial Network (AAAI 2025 · similarity 0.83)
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.
