Translator for American sign language using hand detection and image classification to translate signs into letters. Utilizes media pipe, tensorflow, and keras model to do model training and drawing landmarks for hand detection.
CV2
Numpy
Python 3.6
Tkinter
Tensorflow 2.6
Gui
Mediapipe
spellchecker
All photo data can be produced locally using photographer and take photos by pressing 'q' within the camera frame. There are 28 categories A-Z plus delete and space. You can refer to the link below to reproduce those images basd on their dataset.
https://www.kaggle.com/datasets/danrasband/asl-alphabet-test
To run the interface
python ./mediapipe_solution/app.py
To install dependencies
pip3 install -r requirements.txt
Dylan worked on fixing the gui interface and training the model which mainly involves
- gui.py
- Main.ipynb
Sammy worked on the application and model building / testing as well as data generation which involves
- customdata
- photographer.py
- cnn_testing
Dylan and Sammy both worked on improving the user experience on translating process
- app.py