tl;dr: I want simple hand drawn shapes (boxes, lines, circles) to be identified by my code when the user makes that pattern with their cursor.
I want to go from mouse input to solid objects based on the (rough) shape that is drawn. This is not edge or blob detection, nor vector art as far as I'm aware, but some sort of path direction/line segment detection. There are X preset shapes, and drawing something similar to one of them will create that shape at that size.
In Trine, this includes a square for a box and a line for a plank (not sure if there are others, but I don't mind that being spoiled for me if there are). So far I'm getting *way* more complex stuff than I need with my Google searches, I think, so turning here for more practical applications at my needs level. "Evolutionary visual learning" was described in one document I found, and that's far beyond anything I want.
Working in C#/Unity, in case there are any relevant libraries I should look at.
EDIT: These two assets in Unity seem to do what I want, but I'm looking to understand how it's done in case I can implement it more cheaply/effectively/simply myself for my own purposes (rather than adopting a new engine to what I already have)