I’ve created a new environment for AIRIS that tests its ability to recognize basic drawn characters. I draw a character, show it what to recognize that character as (label it), and it correctly guesses that character when presented with it again. It can also recognize a variety of variants to the trained character.