@galaxykate recognizing sign language or emoji faces from a camera feed are both popular, though 20m probably isn’t enough time to do it from scratch. Maybe with a pre built notebook?