This library is a companion for the French Belgian Sign Language dataset. You will find useful functions to load and manipulate the video from the LSFB dataset. The package provide a pytorch dataset class and several useful transformations methods for video.
A complete documentation for this library is available here
This library is officially deployed on PyPI. This means that, for once (don't lie), the code must be tested and validated before deployment.
This section will explain how to run the test suite, how to set up a local environment enabling you to use the library as if it was installed from PyPi and how to finally deploy when all is good.
In a near future, all the process should be automatised when a pull request is accepted on master.
This project use the tox and pytest for testing. Tox is a tool for running tests in multiple environments while pytest is the most commonly used library for writting test suites. The configuration of the tox environments are located in the tox.ini
file.
To run the test suite, you just have to run the command tox
in the lsfb-dataset directory. If some dependencies were added to the setup.py
file, you need to run tox --recreate
in order to force recreating the test environment including that dependency.
Writing good test is trickier, please refer to the pytest documentation for more information.
The project use mkdocs for its documentation. You need to install the package mkdocs "mkdocstrings[python]" mkdocs-material
to build the doc. The docstring format used in the project is the Google Docstring
You can always install the library locally by running the command python -m pip install ./lsfb-dataset
in the root directory.
If you use this library or the associated dataset, please cite the following paper:
@inproceedings{Fink2021,
doi = {10.1109/ijcnn52387.2021.9534336},
url = {https://doi.org/10.1109/ijcnn52387.2021.9534336},
year = {2021},
month = jul,
publisher = {{IEEE}},
author = {Jerome Fink and Benoit Frenay and Laurence Meurant and Anthony Cleve},
title = {{LSFB}-{CONT} and {LSFB}-{ISOL}: Two New Datasets for Vision-Based Sign Language Recognition},
booktitle = {2021 International Joint Conference on Neural Networks ({IJCNN})}
}