Hey Michael,

How possible would it be to implement the option of generating a quads only mesh? Or an 'almost quads only' mesh I should say (some triangles here and there). Something in the way Groboto or PowerBooleans (3dsMax plugin) does it. Basically a mesh that you could apply subdivision on and it would look 'as expected' (very generic term here).

That would definitely be a key factor!

Thanks!

The day it would be possible to translated 'as expected' into math formula I guess :P

I think this is not possible.... other code etc.

Peace,
Rich_Art. ;-)

Hi Jacobo,

> How possible would it be to implement the option of
> generating a quads only mesh?

Unfortunately no, not without a significant amount of work because creating that kind of mesh involves tiling trimmed surfaces with a lot of little quads in a much different way from the current mesher, and creating a totally new topology where all those rows of quads end up slamming into one another which is the particularly tricky part.

See here for some previous discussion:

http://moi3d.com/forum/index.php?webtag=MOI&msg=4600.1
http://moi3d.com/forum/index.php?webtag=MOI&msg=3628.227
http://moi3d.com/forum/index.php?webtag=MOI&msg=4643.7
http://moi3d.com/forum/index.php?webtag=MOI&msg=3628.227


> Something in the way Groboto or PowerBooleans (3dsMax plugin) does it.

I believe that PowerBooleans uses the same code that's in the Solids++ kernel that MoI uses. I tested it and found it to make messy results most of the time, so that's something that seemed like it would work only for a couple of situation and was not generally that robust.

The Groboto one looks pretty interesting, but if you follow their discussion thread on the Modo forum you will find that they're saying that they always seem to be in the process of reworking and changing it. So it's a large amount of ongoing work that they're doing on that. That should be a good indication on the amount of time that's involved in making that work.

Probably your best bet is to use a retopologizing tool for reworking a mesh into quads, something like TopoGun or 3D Coat also has some good retopo tools in it. Maybe at some point Groboto will deal with imported data from other CAD systems, it would be good to ask them about that.


It is something that I would like to try to tackle some day but it's such a significant amount of work involved that it is difficult to find the huge block of time that would be necessary to do it.


- Michael

I totally understand that. I'm basically in the process of evaluating a few different options for different tasks so I'm gathering information, testing a few things, and then make decisions based on what I found and I think I consider better for our specific needs. I'm just looking for things that could save time, how and under what circumstances.

I was kind of assuming that getting an only quad mesh wouldn't be an easy task. I remember discussing that in the Marvelous Designer forums to get quad meshes out of it. They started implementing one of the different approaches they were researching (some of them suggested in the forum). That was I think like almost two years ago, and it hasn't been implemented yet.
But I didn't know how far you were from that in your software having already what you have implemented, so I thought it wouldn't harm anybody to ask.

Thanks for the links to the other posts, by the way. Sorry if I ask questions that I've been replied before. I try to look in the forums before asking, but most of the time I can't spend time enough to find what I'm looking for. I'm trying to get to a conclusion as quick as I can. I hope you don't mind I ask :)

Hi Jacobo, it's no problem to ask - the forum basically exists for that purpose!

It can be worth a try to do a really quick try using the forum search but if you do not immediately find it then go ahead and ask. I usually will remember some more details about the prior posts and that will help me to do a search with some more specific keywords.

- Michael

Hi Jacobo,

> They started implementing one of the different approaches
> they were researching (some of them suggested in the
> forum). That was I think like almost two years ago, and it
> hasn't been implemented yet.

I would not be surprised if it turned out to be much more difficult than they initially expected.

Right now the primary way that "all quad" meshes are created by people is by arranging edge flows by hand in a sub-d modeling program, using the modeler's judgment and experience in how to best arrange the topology.

It's very difficult to replicate things that are based on that kind of intelligent judgment just with a list of exact instructions like a computer program is made up of.

The process for doing sub-division smoothing is fairly finicky too, it's quite easy to get a lot of messy results (with stuff like lumpy areas) in the subdivided results if the topology is not really well spaced and well formed. So I mean it's not even just having something technically "all quads" that's the problem - the thing you're looking for is more like "all quads with optimal edge flow topology same as hand drawn"...

- Michael

What? Aren't we in 2012 yet? I've grown up with the idea that by now machines would be smarter than humans. Damn TV. I'm disappointed... :)

Seriously now. At this point I'm not looking for an "all quads with optimal edge flow topology same as hand drawn"... button (well, yes, I'm looking for that, but only if I was the only one who knew about it :). In this particular case I'm looking for things that pretty much exist, because I've seen them somewhere else, but what would make the difference for me is to have some of those things combined together in certain way.

Groboto for example generates a mesh far from ideal, or not even valid depending on what you want to do with it later on, but good enough for certain things that I would be doing with it in some cases. If the time saving to model with Groboto instead of in a poly modeling package is worth it, I can deal with that resulting mesh. Best case scenario (on paper), a huge time saver. Again, just in certain situations when that's what I need. It's all very case specific, and there's not one solution for every situation, but that's what I'm trying to evaluate. The same thing applies to Moi...

Thanks again for your support. Very appreciated!

Hi Jacobo, yeah I understand what you're saying but it's not that great of a situation for me to put out something that is too half baked and would generate bad results in a lot (or maybe even the majority?) of cases even if it would be useful in some other kinds of specific special cases as well...

Right now I'm just not confident that I could make an all-quad mechanism that would not generate nasty results really frequently. I think that there would be a huge amount of time involved in making one that is very robust.

It requires such a different approach to forming the mesh that it is not really all that related to the current tessellator.

- Michael

Also meanwhile there are retopology tools out there that are targeted at reworking a mesh to solve this specific problem...

Have you looked at 3D-Coat's auto retopo function?

Maybe taking triangulated output from MoI and taking it through 3D-Coat's auto retopo mechanism would produce some of the "good enough for a particular use case" type result that you're asking about.

- Michael

Hi Michael,

I've already tried 3D Coat (which is a great piece of software), and somebody is already using it in production, but retopo is actually one of the parts of the process that could be avoided in certain situations (insist, in certain situations). I just was curious about what would involve to implement such thing. As I said, I'm evaluating options, trying not to discard them too soon when they don't seem possible, and see how far things can get in certain aspects. Not that "I need that or I won't be able to continue with my life", just gathering information.

Moi with its current exporting to poly features already helps to speed up a lot certain tasks.

Thanks!

Hi Jacobo,

> but retopo is actually one of the parts of the process
> that could be avoided in certain situations (insist, in
> certain situations).

Why though? If it can pump out an all quad mesh for you how and you were saying that an all quad mesh could help you in certain situations, it is a tool that could help you get the kind of output that you are asking about...

Why rule it out if it can get you the type of output that you are requesting here?

- Michael

No, I'm not ruling out anything. As I said, we're already using it. Just considering different options that can help in certain situations. I don't want to be misunderstood. I would need to explain the whole thing to give you an idea of why this would help in certain cases and how compared to other solutions.

It's not a big deal though, just curious.