In this example, we'll use a pretrained MobileNet model and train another model using an internal mobilenet activation to predict up to 5 of different classes from the webcam defined by the user.
Created by Ben Byford, editing the work of the lovely people at Google - original example some code and inspiration from: Google pacman example live!