Press one of the buttons ("Add X" where "X" is "A", "B", or "C"). Each time you click each of the buttons, a single frame from the webcam is added as training example. The model will continue to make predictions ("A", "B", or "C") from the webcam images.