Machine Learning

In order to determine what symbols are being shown in front of the camera, we implemented a convolutional neural net using Keras to classify the kinds of images it sees as a certain letter in ASL. For the algorithm to actually learn what gestures mean what, we gave it over 7,000 images of people signing the ASL alphabet. From this set of training images, it saves similarities and patterns for the same letters so it can then recognize when they come up again in front of the camera.