This project was done in colaboration with Chirag Gomber
A deep learning network trained to identify ASL alphabets from a live video stream. The letters X, Z, 'nothing' and 'cancel' were excluded from this dataset.
A future goal is to add these letters into a word along with an autocorrect feature. The dataset can be obtained from here.
Transfer learning was performed on a ResNet50 network. The model was pre trained on the imagenet database.
The file resnet_model.py contains code used to train the network.
runModel.py runs the network with a video stream from a webcam.
