Excessive triangulation

[Fonte Boa]

First of all, congratulations for MeshMixer: very interesting concept, i would like to test it more.

I am a intensive Wing3D user and got news about MeshMixer at ODF forum.

So my first experiment was to send a mesh from Wings3D to MeshMixer, via OBJ format
(i've already downloaded and installed dec.6 last MM installer, ok)
OBJ export Wings3D dialog allows me to choice between two kinds of Tesselation (Triangulation and Quadrangulation), or either NONE. I could get some serious MeshMixer crashes when no tesselation option was set (if you are interested i can send you some simple meshes that i couldnt load at MM.

But i would like to ask you if at future MeshMixer couldnt avoid to generate so many triangles and try to minimize this aspect. I really work will (almost) all quad meshes, and the main issue i could experiment with MeshMixer is its tesselation algorithm. There are situations at which if i just export an OBJ to MM from Wings, load it, save it back from MM and load it again at Wings, i can see unecessary tris...

http://www.youtube.com/watch?v=9lDKeBF_E_M

Anyway, congratulations for the great work and excuse my bad english.

[RMS]

Yes, please send me meshes that will not open. Also if you could send me that example you use in the video (the part file too, if you can), that would be helpful.

I will try to improve the triangulation, but it is *very* difficult to stitch with quads...
cheers,
-RMS

[Fonte Boa]

Thanks for the reply.

<<it is *very* difficult to stitch with quads>>

I imagine so. Between when original mesh was built all-quads, maybe you can code MeshMixer so that it can "find" coincident verts and just weld them, as an alternative to  fill the boundary region with triangles.  Then the retopology work time (at another 3d app) can be minimized.

Maybe if ROI Scale can be set to 0 (=zero) (no actually possible)
and if cut part could keep its original scale
and if MeshMixer could, under this situation, just weld coincident vertices (*)
(and/or allowing vertices with 3 edges, 3-poles ones),
maybe undesired triangulation could be avoided.

(just thinking aloud)

Thanks for your attention.

EDIT:
(*) or forcing this weld operation.

[RMS]

Yes, I have thought about welding. I could relatively easily collapse the triangles I add, but the pastes would not be as smooth. Do you think that is less important?

cheers,
-RMS

[Fonte Boa]

<<I could relatively easily collapse the triangles I add, but the pastes would not be as smooth. Do you think that is less important?>>

If your are managing all-quad mesh, the most important would be to keep edge loop flow.
This probably isnt so easy to do, but any effort to avoid triangles at boundarie region would be very appreciated.
Because, anyway, animators that can make use of MeshMixer
will probably re-topology that region (to keep the results under control)
and will spend time in this.

(currently i will begin to test if a reazonable workflow could be
export the mesh all triangulated to MM to then get it back and just try do untriangulate it:
i dont expect great results from this, but i have to try)

The ideal would be to keep that region all quad,
either by collapsing the triangles even by leaving 3 edges poles.
A tolerance setup to (automactly?) weld verts would be useful in this case.
Or either a second alternative:
1st one: fill with triangles
2nd one: keep quads, respecting edge loops...

(i am not sure, but there can be users to who those triangles dont matter:
to animators that prefers to work with quadraculate tesselations,
any alternative would pe really appreciated).

tks for your attention, RMS.

[Dimitri]

Agree with Fonte Boa... it would be of a very essential importance having, at least, a way of producing a final  quadrangle polygon mesh after the 'sculpting's having been finished.

I do not know how possible is, with the present programming ways we have, to take a so densely triangulated mesh and convert it to a quadrangle polygon mesh so that it may be exported in a clean cut shape, in obj format, for example. But the concept is very easy to be explained and understood.

Sculpting with the method of triangulations gives very nice results surely, but there remain the problem of having clean cut 'loops' on a mesh so that it may be edited further as a polygon object (or as an object to be animated).

Having not  a way of achieving such a thing would make any such sculpting app remain as an interesting 'toy', somehow, at most. And it would be a pity really.

[RMS]

I agree that this would be a nice feature. It is very difficult to do automatically in a nice way. The tool 3DCoat has a method for automating, but still I think it does not give enough control over the loops.

There are lots of other tools you could use for this task - I know that at least zbrush, modo, 3DSMax all have tools for drawing a new topology on an existing mesh. Topogun also does this, I think.

I have ideas for advanced versions of such re-topology tools, but they are not going to be coming particularly soon.

[Dimitri]

RMS, really, just this is the needed thing: an automatized retopologizing of the whole object, so that there may appear a new object, meshed in quadrangles, ready to be exported as an obj file, for further editing or animation.

[Dimitri]

Here is something related to the subject:

http://www.zbrushcentral.com/showthread.php?t=96408&page=1&pp=15

[hieveryone]

It is very difficult to do automatically in a nice way. The tool 3DCoat has a method for automating, but still I think it does not give enough control over the loops.

I used the trial of 3dcoat just to try the autopo function. It's interesting, but as you say it doesn't give enough control, it still creates some n-gons and you have to use many faces bacause the way it manages small areas like fingers with a low poly count is not good; still I have to admit that overall it does an amazing job. As for the image DrPetter have shown, as interesting as it can be it's still a wip and it must be considered as such.
Retopology is a tricky subject because the way it's done depends on what you need that topology for. I think that if meshmixer should be able to do retopology it would be better to have some solid and easy tool to make the work manually more than an autopo function, but that's just my opinion.

[RMS]

I would rather do a good assisted-interactive tool, then an automatic one. I have plans, just not the time =)