Hi all,

I have seen some very nice demo where you can draw the outline of a scenery, cat or handbag and it creates a (more or less) realistic picture following your shapes.

How much work would it be to generate pictures of planets? you select which parts are water, forests, mountains, craters and it generates something like this: https://images.app.goo.gl/76jbnxVDg8YFE64g6

How much technical and computational work would this involve? I know very little about GANs, but I’m familiar with python (I study reinforcement learning) and have access to beefy GPUs.

Actually ideally I’d want to generate 3D models of the planets, but as a first approximation a 2D image would be enough.