You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
sbb_binarization/README.md

70 lines
2.2 KiB
Markdown

4 years ago
# Binarization
4 years ago
> Binarization for document images
4 years ago
1 year ago
[![pip release](https://img.shields.io/pypi/v/sbb-binarization.svg)](https://pypi.org/project/sbb-binarization/)
[![CircleCI test](https://circleci.com/gh/qurator-spk/sbb_binarization.svg?style=svg)](https://circleci.com/gh/qurator-spk/sbb_binarization)
[![GHAction test](https://github.com/qurator-spk/sbb_binarization/actions/workflows/test.yml/badge.svg)](https://github.com/qurator-spk/sbb_binarization/actions/workflows/test.yml)
4 years ago
## Examples
4 years ago
<img src="https://user-images.githubusercontent.com/952378/63592437-e433e400-c5b1-11e9-9c2d-889c6e93d748.jpg" width="180"><img src="https://user-images.githubusercontent.com/952378/63592435-e433e400-c5b1-11e9-88e4-3e441b61fa67.jpg" width="180"><img src="https://user-images.githubusercontent.com/952378/63592440-e4cc7a80-c5b1-11e9-8964-2cd1b22c87be.jpg" width="220"><img src="https://user-images.githubusercontent.com/952378/63592438-e4cc7a80-c5b1-11e9-86dc-a9e9f8555422.jpg" width="220">
4 years ago
## Introduction
This tool performs document image binarization using a trained ResNet50-UNet model.
4 years ago
4 years ago
## Installation
Clone the repository, enter it and run
`pip install .`
4 years ago
4 years ago
### Models
Pre-trained models in HDF5 format can be downloaded from here:
4 years ago
https://qurator-data.de/sbb_binarization/
4 years ago
1 year ago
We also provide models in Tensorflow SavedModel format via Huggingface and Github release assets:
https://huggingface.co/SBB/sbb_binarization
1 year ago
https://github.com/qurator-spk/sbb_binarization/releases
With [OCR-D](https://ocr-d.de/), you can use the [Resource Manager](Tensorflow SavedModel) to deploy models, e.g.
ocrd resmgr download ocrd-sbb-binarize "*"
## Usage
```sh
sbb_binarize \
-m <path to directory containing model files \
<input image> \
<output image>
```
Images containing a lot of border noise (black pixels) should be cropped beforehand to improve the quality of results.
### Example
sbb_binarize -m /path/to/model/ myimage.tif myimage-bin.tif
To use the [OCR-D](https://ocr-d.de/en/spec/cli) interface:
ocrd-sbb-binarize -I INPUT_FILE_GRP -O OCR-D-IMG-BIN -P model default
## Testing
For simple smoke tests, the following will
- download models
- download test data
- run the OCR-D wrapper (on page and region level):
1 year ago
make model
make test