ATR training data generator

This script downloads pages with transcriptions from Arkindex and converts data to ATR format. It also generates reproducible train, val and test splits.

A documentation is available at https://teklia.gitlab.io/atr/data-generator/.

Installation

• Use a virtualenv (e.g. with virtualenvwrapper mkvirtualenv -a . atr-data-gen)
• Install atr-data-generator as a package from a local clone:
  1. The teklia-document-processing library is setup via git submodule. Please run git submodule update --init.
  2. Then install both packages via pip install ./document-processing -e .