Full Version: A free subd to NURBS solution

Show messages:  1-17  18-37  38-48

// Maybe if you ask PaQ you will be provided with much more "heavy-duty" and reliable stuff :)

Hi Marco,

Simple example like the one you provide is better ... because I didn't realize so far that there was a continuity problem :/
In fact when dealing with super simple shape, I tend to add one or 2 level of catmull clark subdivision prior the conversion, the get a shape that looks closer to the one I have in Modo.
(It`s also a lame method to get 100% quads, whatever the topology was made of)

It seems to greatly reduce the continuity problem Michael describe (but not remove it completely) ... in the other hand surfaces around valence 5 vertex looks less `smooth`.

That having said, is there any existing tool around that gives a perfect result for this kind of conversion ?
Because from memory, T-spline (fusion 360) wasn't giving much better surfaces around poles either.

Thank you Max for the update, I obviously miss (or forgot) you previous message on the subject.

I get the idea of an external tool for this job, but in the other hand, Moi has already the best mesher, having the best sds to nurbs converter would add lot's of attraction !
... but maybe it required to much R&D for a single person.

---

Hello PaQ!

@You : "...I tend to add one or 2 level of catmull clark...It seems to greatly reduce the continuity problem Michael describe..."

Thanks for the very useful tip!

@You : "...... but maybe it required to much R&D for a single person..."

I think it's nearly IMPOSSIBLE for a single person company.
Some sort of collaboration would be absolutely necessary, but, clearly, this would complicate the "trade matters".

Ciao!

---

Hi guys,

I love reading about this subject. As part of my work involves creating toy models, I'm always searching for the perfect bridge between polygons and solids.

Max's solution for MoI is a great one, I am very happy it's here, and grateful that it's free, although I would happily pay extra for it.

The only drawbacks are that you need an all-quad and fairly low-polygon topology for the conversion to work out neatly. Better surface continuity for triangular polygons would be fantastic, and so would it be if the result could be optimized to as few patches as possible, based on curvature, so you could convert high-polygon models without a heavy, cluttered conversion result.

---

@Metin : "...I would happily pay extra for it..."

+1
I too!

---

@PaQ

Hi PaQ,

maybe someone can compare Marco's test model with this solution,
to see if there is a difference:

http://www.resurf3d.com/Objmesh2solid.htm

(The algorithm used in this software is based on the paper named Constructing C1 Surfaces of Arbitrary Topology using Biquadratic and Cicubic Splines written by Professor Jorg Peter.)

Regards
Stefan

---

I googled over a dozen papers with J Peters name.
- B

---

They are linking to this paper:

https://www.cise.ufl.edu/research/SurfLab/pre99-papers/9294.pdf

@Max

>I'll think about this idea

Thanks Max!

Regards
Stefan

---

Hi guys,

The Resurf application is very old, and looking at the examples it doesn't do anything other than what Max's tools already do in MoI — converting each polygon of a low-res quad mesh into NURBS surfaces to form a solid.

Being able to do this in MoI would make me a veeery happy man:

Video 1:
https://www.youtube.com/watch?v=sPB-bHbKsJk

Video 2:
https://www.youtube.com/watch?v=4rDQCxGaLZ8

---

Hi Metin,

sure it is very old, but does it produce the same little errors
Michael has shown with Zebra Analysis on Marco's model?

That was the question... ;-)

Best regards
Stefan

---

Hello Max and everyone.

I've assembled another (I hope) useful model to test the SubDiv scripts and its results.

I've made the base OBJs with Rocket3F and exported to Moi V3. Then applied the Max's SubDiv script.

Here's the 3DM file : http://take.ms/H7CMS

Here's the "Obj_Level_0.obj" : http://take.ms/yXXj4

Here's the "Obj_Level_1.obj" : http://take.ms/53rjXL

Please keep in mind that "Level_0" means that no subdivision was applied to the original obj model.
"Level_1" means that 1 level of Catmull-Clark subdivision was applied with Rocket3F to the obj model before exporting the obj to Moi.

In Moi I've used these settings to perform Max's SubDiv :



Now I don't have enough spare time to do an exhaustive checking of the subDiv results,
but you can use these stuff to make a test on your own, even better if you ca do a zebra analysis with a tool like Rhino or other.

As you can see within this 3dm file there are different "difficult" zones, different "criticality" areas for the SubDiv process.

Here's a Moi's screenshot :



I hope that this can help.

Have a nice day :)

---

Hi everyone I would like to resume this discussion; the difference I noticed, using FreeShip to convert subdivision surfaces in addition to the extreme speed in generating nurbs surfaces, is also sometimes a clean and simple surface, and in some cases follows the edges of the object.
Then you could in moi3d select the edges of the surface to be sharp or to use as a limit of the surfaces a bit like when using the edges creases in Blender, before generation of surface.
A last one also gcad3d converts files in vrml1 format if someone want try my tips..
Below the link to the file that generates the nurbs but I'm not sure ..:P

https://github.com/markmal/freeship-plus-in-lazarus/blob/5774a88d1e050b7eb58d7c2a81dc85e94f8b6768/Units/FreeIGESUnit.pas

Here a new test with the difference on patch generation, but even this is a bit mess, and will be awesome have the ability to determinate the edge of resulting surface even manually with some work flow on the script. But anyway the Max's script is very useful and that's one of the reasons I bought moi3d, which also works in Linux, and now I have a good CAD modeler for some task if someone want a done in Nubrs, but I'm too familiar with Blender where I do every thing also CAD design and sheet metal parts.

Image Attachments:
sub_surf_to_nurbs.png 

---

Show messages:  1-17  18-37  38-48