This system does what your describing The Lanier Verbatim Memory System | Part One: Intro & Part Two: How It Works
Also the ‘cube within a cube’ approach isn’t optimal because your spatial locations need to be distinguishable in some way in order for them to be memorable. For instance you can have multiple swamps but you should try to give a unique element to each one (and possibly even a name) to help you distinguish the locations.
As far as the ‘people learn to play blind fold chess’ I would agree. There are plenty of examples of people improving there visual memory (architects, artist learning to draw from memory, chess.)
That being said, I would recommend not getting too obsessed with trying to intentionally train your visual memory. if you just get used to thinking in images and make a continuous conscious effort to see your images clearer and clearer then you should get better.