PDA

View Full Version : Discussion on Triangles in Organic Meshes


spiggy
07-16-2005, 04:46 PM
Okay, so here is my understanding of triangles in an organic polygon mesh:

It is a fact that most high-end organic modelers, such as Jeff Unay, donít consider the amount of tris they create until clean up stage further down the modeling process. Albeit in saying, they donít let the tri creation slow them down in blocking and sculpting the required edge loops in the intermediate stages of building an organic model.

If left untreated, triangles, when deformed, can cause unexpected results on a surface, however, in saying this, triangles are necessary in the modeling process to be used in strategic places.
They are used, generally, to discontinue edge loops (to stop them from going all the way through a character, hence to localise muscle regions)and to change directions in edge loops for muscle flow and deformation purposes. Going on from many arguments the conclusive decision is: that when cleaning up a mesh, if triangles are necessary, then modelers should hide them in concealed places, such as (with humanoid characters) in the armpits, in-between the clavicle in the spine, in the calf, in the hair, in the gums or at the sole of the feet etc.

This made me think though. When are triangles truly necessary?

If, for instance, most organic beings have 2 biceps, 2 triceps, 2 pecs, 2 quads etc, etc, then these triangles can be chased across the meshes topology to form a quad.

So in what instances should tris be necessary?

My conclusion is:

a.) With asymmetrical meshes

and

b.) When time constraints/deadlines are impractical to clean up the mesh to that extent. I.E the clever cheats way.


Admittedly I havenít given this the deepest amount of thought, soÖ

What are your views?

Yourworstnightmare
07-16-2005, 06:20 PM
Check out this thread, it's almost 100 pgs all about topology.

www.cgtalk.com/showthread.php?t=108412 (http://www.cgtalk.com/showthread.php?t=108412)

On a face triangles are a no no, but on a body where not everything is symetrical triangles can actually be beneficial because of the creases on the human body can be made quite easily with correctly positioned triangles.

Hope i was of some help.

gardogg
07-16-2005, 10:36 PM
I have found that there is always a way to eliminate triangles. The more one practices at getting rid of them the quicker and easier it becomes. The geometry really looks better once every triangle is eliminated. I think that in general, one should avoid triangles. Its a challenge and a good habit. Unless of course there is a time contraint. Then I guess leaving a few won't hurt.

spiggy
07-17-2005, 01:43 AM
You certainly weren't my worst nightmare!


Cool, good to find a benefit for tris!

Opelfruits
07-17-2005, 09:42 AM
i see no problems with using triangles, the final fantasy characters from the film use triangles, and lots of them, even in places where lots of deformation is happening.

if they dont effect the smoothness of the model then what is the big problem everyone has with using them. I would rather concentrate on the way the model looks than trying to eliminate every triangle.

tibes
07-17-2005, 09:48 AM
i see no problems with using triangles, the final fantasy characters from the film use triangles, and lots of them, even in places where lots of deformation is happening.

if they dont effect the smoothness of the model then what is the big problem everyone has with using them. I would rather concentrate on the way the model looks than trying to eliminate every triangle.

I agree. While it's always possible to eliminate triangles, it's not always possible to eliminate triangles without increasing the density of the mesh. Personally I don't think there should be any hard and fast "rules" in modelling, you should just approach each situation as best you can. If triangles cause a problem, remove them, if they don't, then spend your time somewhere else.

gardogg
07-17-2005, 08:55 PM
Lets say your posting some wireframes of your models and they've got triangles here and there. Wouldn't people think that it looks careless or sloppy? The arguments above in favor of using triangles are convincing to me. Now I don't know what to think... Do I avoid them or not?

spiggy
07-17-2005, 09:13 PM
good point Gardogg! I.E Especially from a professional POV.

Opelfruits
07-17-2005, 10:22 PM
when should you avoid triangles - when the mesh will smooth badly because of the triangle

when you can use triangles - when it wont effect the smoothing

the main thing is that the model looks good, and if it needs to be deformed that it deforms properly. If triangles stop you from achieving this then dont use them, if they aid you then do use them.

tibes
07-18-2005, 12:52 AM
If people give you a hard time about triangles in your mesh, they're probably just looking for something to critic you with. Call them on it. Wireframes are useful for criticing someone's edge flow or mesh density, but the real test is how the object looks shaded, cause no one's gonna look at the wireframe when you've rendered it out.

Again, it looks good, don't worry about it. :)

RoundRobbin
07-18-2005, 07:42 AM
nah, tri's are a no no. I hear most renderers don't know what to do when their confronted with a tri, so they render it half ass and the result isn't as good as a four sided face. Most experienced modelers who get lazy make tri's so they could get a crease, usually tho, that crease doesn't look appealing in the model. however, if the model is near finished and the tri is in an un-important area, where it doesn't get deformed and the camera doesn't see it than just 4get it, it's not that big a deal.

tibes
07-18-2005, 02:08 PM
I hear most renderers don't know what to do when their confronted with a tri, so they render it half ass and the result isn't as good as a four sided face

I'd like to see an example of that. Don't always believe what you "hear". I've never had problems like that in my experiences. Many tris in a sub-divided mesh will become quads anyhow. Some (not all) renders convert quads to tris at render time, and pretty much ALL game engines do. If tris were that much of a problem then this wouldn't be the case.

Personally I find that poles can be more a problem for smooth meshes, and poles tends to happen a lot in all quad meshes where edges change direction.

It seems like the "no tris!" rule has just perpetuated from teacher to student and not so many people have bothered to think about *why* that piece of advice was given. The no tris rule is a good guideline which encourages good habits, but I don't think it was ever intended to be gospel. Like I said before, you need to assess whether improvements are necessary based on how the model looks, and whether that area needs to be animated, and if it will be, how it will deform.

:shrug:

pauljs75
07-20-2005, 08:26 AM
I hear most renderers don't know what to do when their confronted with a tri, so they render it half ass and the result isn't as good as a four sided face.

Sounds like you've confused tris with n-gons. Quite a big difference between the two, and believe it or not - some programs convert everything to tris before rendering. You'll notice that especially if you have some non-planar quads.

As for when to use tris? Use them when the shape of the mesh requires it. With some types of curved topology it's pretty hard to avoid them. Also it's not just tris that may cause problems with mesh deformation but poles/stars (where many edges connect to a single vertice.) Most people (myself included) generally like to work with quads because it's a bit easier to follow the topology. Also if you use edge loops and rings when developing a mesh, quads tend to work better. (You can jump from one loop to the next, and not have to worry about how they terminate, etc.)

lostlittleboy
07-20-2005, 05:02 PM
nah, tri's are a no no. I hear most renderers don't know what to do when their confronted with a tri, so they render it half ass and the result isn't as good as a four sided face.

Eh? :curious: (http://misc.php?do=getsmilies&wysiwyg=1&forumid=0#)

Its not just some programs, its pretty much *every* renderer, be it a games engine or scan-line renderer or whatever. (I would say 100% every renderer, but I wanna cover my arse on this one! :D) NURBs, sub-ds, 4 sided polys, n-gons, they all get translated into tris when they're rendered.

Cessen
07-26-2005, 10:44 PM
nah, tri's are a no no. I hear most renderers don't know what to do when their confronted with a tri, so they render it half ass and the result isn't as good as a four sided face.

No, that's not true. Every renderer in existence that renders quads can render triangles just as well. In fact, triangles are easier to render in a lot of respects because they are guaranteed to be flat, whereas quads (at least in character modeling) usually aren't flat and actually represent a curved surface. Many renderers actually split quads into two triangles before rendering.

Sounds like you've confused tris with n-gons.

Yup. It's n-gons that renderers have trouble with, because there isn't a good way (yet, anyway) for them to decide how to tesselate them into quads or triangles before rendering. Which vertices do they draw edges between? How will that effect shading?

Also, I'd like to point out that whether or not a renderer can "handle" triangles or n-gons is a moot point with catmull-clark subdivision surfaces since both triangles and n-gons alike are turned into quads with the very first subdivision. So in that case it's all about how they affect the shape of the subdivision.

CGTalk Moderation
07-26-2005, 10:45 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.