Pre-print: https://arxiv.org/abs/2404.07211 Web application: https://signlanguagerecognition.aprendeconeli.com/
Licence Attribution-NonCommercial-ShareAlike 4.0 International: https://github.com/ecabestadistica/SignLanguageRecognition/tree/master?tab=License-1-ov-file#readme
notebooks for training the models
datasets for training the models
Datasets currently at: https://drive.google.com/drive/folders/1oWIGEy2vciJK8Wfl79Xg64JffZbjmzfY
pretrained models just for testing
Trained models currently at: https://drive.google.com/drive/folders/1u_H4_2Wc2-plryjFQnZ9C52HyTXHGn-5
the apps/scripts for testing the models
Example images for testing app
simple web server for testing the models from anywhere without need to install the environment
Note About the dependencies, tensorflow 2.14.1 works with cuda 11 if you have cuda 12 you need to install tensorflow 2.15.X also you need to take care of tensorrt to make sure is compatible with cudnn
This folder constains Pipenv file for installing the dependencies. To install the dependencies, run the following command:
pip install pipenv
pipenv install
To run the app, run the following command:
pipenv run python testApp/test_app.py
This folder constains Pipenv file for installing the dependencies. To install the dependencies, run the following command:
pip install pipenv
pipenv install
Configure your jupyter to use the pipenv kernel and then run the notebooks in the notebooks folder. Follow the instructions in the notebooks to train the models.