You still need some smarts, since the picture you just took won't be in the training data.