
Segment geometry
What is the most Houdini way to do this nowadays
By segment, I mean the process of partitioning an image into distinct regions or pixel-level masks and linking them to natural language prompts or semantic concepts
-- assuming this is done on a rasterized version of the curve set, and that the regions attributes are then transferred back to the original curve afterwards.
For example: starting from an EPS file of a side view of car, I would like to get some kind of attribute for each point / curve / connected piece, in order to make it easier to make something with said curve.
Any thoughts ? There is so much data available in the context of AI training, it would be nice to be able to augment any geometry in such a way. I guess it's already solved in a much bigger context though, as is almost anything nowadays
For example say you have this in SOP as curves.
how to segment it semantically, to help rebuild it as a 3d model, set texures and whatnot