Comment by dthunt

Probably not training "matching", given that there is no place where you indicate what the correct thing was.

It could be gathering data about how people draw shapes in general, but it's not immediately obvious to me how much that can help.