This model was made under a day for the TGI Hackathon and trained using about 180K images of American Sign Language alphabet images which won my team 2nd place.
To just look at the dataset or to train your own model go here!
See our Code at Github.