This project is aimed at training a convolutional Neural Network in order to classify the american sign language fingerspelling. The pictures used on this project were downloaded on kaggle.com. The dataset consists of 87000 pictures for the training set and 29 pictures for the test set. An accuracy of 99.9% was achieved both during training and validation. The model was trained on a Macbook Air with apple silicon (M1) processor.
git clone https://github.com/Huntonion/asl-cnn
download the dataset from https://www.kaggle.com/datasets/grassknoted/asl-alphabet and extract the archive into the asl-cnn folder.
then run:
pip install -r requirements.txt
jupyter lab model.ipynb
if you are interested in knowing more about this project, please consider reading the paper produced.