What it does
Trained on 87.5k images from the Kaggle dataset. It can successfully classify all the alphabets of the ASL with an accuracy over 98%.
How we built it
Leveraging fastai and Kaggle dataset
Challenges we ran into
Augmenting images from individual batches to increase accuracy.
Accomplishments that we're proud of
A very high accuracy using resnet50 architecture!
What we learned
Data augmentation and how it affects accuracy.
What's next for Sign Language Translator
Try to deploy the model.