---
tags:
- image-to-image
license: apache-2.0
---
# About `sbb_binarization`

This is a CNN model for document image binarization. It can be 
used to convert all pixels in a color or grayscale document image 
to only black or white pixels. The main aim is to improve the 
contrast between foreground (text) and background (paper) for 
purposes of OCR. The model is based on a `ResNet50-Unet` model.

# Results
In the *DocEng’2021 Time-Quality Binarization Competition* 
([paper](https://dib.cin.ufpe.br/docs/DocEng21_bin_competition_report.pdf)), 
the model ranked 12 times under the top 8 of 63 methods, winning 2 tasks.
 
In the *ICDAR 2021 Competition on Time-Quality Document Image 
Binarization* ([paper](https://dib.cin.ufpe.br/docs/papers/ICDAR2021-TQDIB_final_published.pdf)), 
the model ranked 2 times under the top 20 of 61 methods, winning 1 task.

For details, see [sbb_binarization](https://github.com/qurator-spk/sbb_binarization) on GitHub.

# Weights
We provide a `saved model` for Tensorflow2.

| Model | Downloads
| -------------| ------------------------
| `2021_03_09` | [`saved_model`](https://huggingface.co/SBB/sbb_binarization/tree/main/saved_model)