SignLanguageRecognition

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

Folder structure

Notebooks

notebooks for training the models

Datasets

datasets for training the models

Datasets currently at: https://drive.google.com/drive/folders/1oWIGEy2vciJK8Wfl79Xg64JffZbjmzfY

TrainedModels

pretrained models just for testing

Trained models currently at: https://drive.google.com/drive/folders/1u_H4_2Wc2-plryjFQnZ9C52HyTXHGn-5

testApp

the apps/scripts for testing the models

testApp/Signos XX,AlphabetASL,Signos Numeros

Example images for testing app

testWeb

simple web server for testing the models from anywhere without need to install the environment

Note if you are in google colab pipenv is mostly not needed

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

How to run the app

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

How to train the models

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.