Install

Calamari requires:

Python3.7 or later including the development packages.
(optional) cuda/cudnn libs for GPU support, see tensorflow for the versions which are required/compatible.

Calamari was tested on Linux, but should be flawlessly usable on windows or mac.

See also the install instructions for tfaip and tensorflow.

Installation using Pip

The suggested method is to install calamari into a virtual environment using pip:

virtualenv -p python3 PATH_TO_VENV_DIR # (e.g. virtualenv -p python3 calamari_venv)
source PATH_TO_VENV_DIR/bin/activate
pip install calamari-ocr

which will install Calamari and all of its dependencies.

To install the package without a virtual environment simply run

To install the package from its source, download the source code and install it. Optionally (but recommended) install in a virtual env.

git clone https://github.com/calamari-OCR/calamari
cd calamari
python setup.py install

Conda users can alternatively call

conda env create -f environment_master.yml

Calamari uses black code styling. It is recommended to integrate running black as pre-commit hook. The following guide helps to setup everything.

The requirements pre-commit and black are already part of the requirements. Setup pre-commit by calling:

pre-commit install

To upgrade the pre-commit packages call

pre-commit autoupdate