Scan&Solve on STL Model of Free-Form Sculpture

Comment

Comment by Armido Cremaschi on November 13, 2010 at 3:04pm

Understood, thanks.
Armido

Comment by Vadim Shapiro on November 12, 2010 at 1:22pm

The notion of "valid mesh" is not specific to S&S. Basically, the mesh must be a valid boundary of a solid, so that integration results are correct. In math jargon, the mesh must be an "orientable manifold" surface. This translates into a set of practical conditions described on page 2 of this FAQ document
Actually, the same conditions apply to any "valid closed polysurface".

Comment by Armido Cremaschi on November 12, 2010 at 4:49am

Thanks, Vadim. Keep up the good work. Would you mind explaining in some detail what are the features of a "valid" S&s mesh?
Ciao
Armido

Comment by Vadim Shapiro on November 12, 2010 at 4:14am

Ah, this is a great question! In fact, we indeed have a version of Scan&Solve that would work directly from a mesh or collection of meshes, or even collection of some bad meshes -- because Scan&Solve procedure is intrinsically robust with respect to small errors. But ... garbage in, garbage out ... if we cannot quantify the validity of computations, we have no idea whether the computed answers mean anything reasonable. Until we know how to quantify these errors, Scan&Solve has two choices: (1) either assume responsibility for the quality of user's meshes, perhaps repairing them as needed, OR (2) require user to put these meshes in something that is guaranteed to be valid. Obviously (2) is easier and users can do it NOW.

You are also correct that, in principle, Scan&Solve could work directly from a valid mesh, without converting it to a polysurface. But for us this would mean maintaining two different versions of Scan&Solve code -- and right now we simply do not have the resources to do so.

Finally, we do envision that future versions of Scan&Solve would work from a collection of meshes/surfaces, but as you can see from answers above, the task is more delicate than it may appear.

Comment by Armido Cremaschi on November 12, 2010 at 3:26am

The report documentation feature is quite useful. Thanks. Most times, my objects are not easily converted into valid polysurfaces, due to their complexity and to the fact that they originate as meshes. Nonetheless, the analysis carried out on the mesh, provided that it is valid, is usually being completed. Would you mind explaining why, having a mesh, it is necessary to create a complex polysurface for analysis? Is it not true that it is anyway transformed again into a suitable mesh? Would it be possible in the future to select a bunch of interacting /connected (poly)surfaces and meshes as input of the analisys?
Thanks.