Our library prepares diverse, and geometrically accurate visual data from generative models to allow robots to learn complex tasks, without depth sensors.