In this project I have used the dataset available at https://www.kaggle.com/datamunge/sign-language-mnist
Although this dataset can be trained using a standard CNN model, In this project I have used siamese network model to recognize american sign language.
Framework used :
Tensorflow, Keras
ASL Dataset Example:
Model used :
Here, Input Layer is defined as follows :
Classification Report Generated on TestSet:
This model is successfully able to classify static sign language image into 24 classes with 100% accuracy on test set
This is because the sign for J and Z are in motion, which i will look to cover in a future project.