The AI research I would love to see (someone has to be working on this) is Lidar point cloud to Siemens x_t b-rep parasolid (or STEP). I a functional proof of concept to 2d projections as DXF (essentially where a point source strikes a surface and extrapolate lines from that). But I feel like going all the way to parasolid may almost be feasible now. You can trick GPT into produce som STL drawings. Anyone know of any projects/research happening in this space?

This is from an AVP app called Magic Room. It would be awesome to go from this to a full fledged capture. Until now I feel this has been WAY out of reach. But now I feel like it is totally feasible to achievable where the video and lidar data are coupled to the input of the model.

Output of proof of concept:
Input: an STL

Output: https://www.printables.com/model/2671...