Tables for 3D production

Become a member of the CGSociety

Connect, Share, and Learn with our Large Growing CG Art Community. It's Free!

REPLY TO THREAD
 
Thread Tools Search this Thread Display Modes
Old 10 October 2009   #31
Originally Posted by mister3d: If you take so much time to make such charts, why not to read what others did? I'm not sure if you got my email answer where I listed this one. In the end it's not my invention either, I just organised what others did. But maybe you will come up with a better presentation or explanation. I'm not sure if anything new can be found in this area. But what is really missed is using tris and n-gones. Not avioiding, but using.
Good luck buddy, I have a high respect for what you do, as I know what it takes, don't get me wrong.

I think this kind of information still needs a good solid dvd, alongside with sub-d modeling tecniques and tips and tricks.


Glad you like it guys.

@mister3d. Yes like i said i need to be more patient to read. I went through college and i rarely investigated stuff, i was always creating. It is a flaw of my character :P
But even looking at what others do, i still have to do my own "charts", because each person organizes information diferently. And sometimes i find SomeArtist and Tonjee information not very pleasent to read or to slot in my head. And since im just looking at it the understandin will be shallow.
Duly noted on the Tris and N-Gons.

Cheers
 
Old 10 October 2009   #32
Another topology thread promoting triangle shaped quads in the same space where tris are discouraged might just be what's needed to send me on a killing spree...
Why not in the modelling forums btw, since this is pretty modelling specific?

As for something related to the original topic: I tink it's a pretty decent idea actually, and the proposed format looks readable and accessible to me.
__________________
Come, Join the Cult http://www.cultofrig.com - Rigging from First Principles
 
Old 10 October 2009   #33
Originally Posted by ThE_JacO: Another topology thread promoting triangle shaped quads in the same space where tris are discouraged might just be what's needed to send me on a killing spree...
Why not in the modelling forums btw, since this is pretty modelling specific?

As for something related to the original topic: I tink it's a pretty decent idea actually, and the proposed format looks readable and accessible to me.



Yea I have to agree on both points, I think the charts are a great idea Very well designed and represented information.

I do have issues with a little of the content particularly about 4point triangles versus triangles. If you look at the Central and Grid divisions for a four point mesh both have 4 point triangles as the solution and both will make terrible geometry when subdivided with catmull-clark. I can't agree to either of those two, really you should never need two 4 point triangles sharing two edges like that, it is never a better mesh, and each subdivision those two triangles get will always produce two more 4 point triangles, where standard triangles will subdivide into nice clean quad meshes. Those points don't subdivide well when it is a curved surface either, they will always lead to a sharp point in the mesh, and make deformations and skinning much more difficult.

Your Lighting/Shading TD's will hate you, your Rigging TD's will hate you In general not a good idea.
__________________
Quote: "Until you do what you believe in, how do you know whether you believe in it or not?" -Leo Tolstoy
Kai Pedersen
 
Old 10 October 2009   #34
Ok, here is the thing. The Table altough follows the rule "All Quads when possible" it does not say do this instead of that.

It's what stated there. "Different Mesh configurations within the same boundary". This because i started to notice that focusing on the outward points would be a good base to construct more complex configs.

So for example, the thing you point out, LucentDreams, about me having all those FourPointTris.That's there exaclty so one knows that all those polys can become a quad.
So sometimes when cleaning up areas of a mesh i had trouble to understand what a set of polygons could become until i actually edit them and see that wasn't what i wanted.

So for example the 6 column, 6 points its what it take to make 2 adjecent quads, 3 quads sharing a vertex of the base config for an X Loop.

It was more of something like; Pick a nifty Mesh config you like, look at the boundary points number and it will be probably more easy to remember it and construct it than remember all the internal config.

Still, all of these stuff i show up here are propositions and open to discussion.

Cheers

Last edited by animatics : 10 October 2009 at 09:21 PM.
 
Old 10 October 2009   #35
Without going into the very large number of reasons why triangle shaped quads are bad, there is one that is pretty hard to argue:
A triangle shaped quad (or any non convex shape or edges being aligned) will royally mess up barycentric coordinates for that polygon, and even if you're only nudging them out a bit, the potential for deformations to push them over the edge (no pun intended) and messing it up for you is pretty high.

That alone is a good enough reason to never even consider it a valid topological choice, and the people promoting it as a "trick" (seen that and even seen it callef a nifty trick) simply don't know how lethal it can be.

In other, more rare, cases like a spring network being built from the cage (cloth, SBD, relax deformers) it will be even more damaging, as the shortest transversal spring will immediately collapse that polygon, and the other transversal spring will be outside the footprint of it, making it glitchy to impossible to solve.
__________________
Come, Join the Cult http://www.cultofrig.com - Rigging from First Principles
 
Old 10 October 2009   #36
So I want to discuss triangles and n-gons.
All-quads technique is predictable, but sometimes making all quads requires more polygons than needed imo.
Using triangles and n-gon's all-throughout is not a good idea, but in some places is quite ok, once again, in my opinion.
Theoretically the drawbacks of using n-gons is a texture stretching, though I can't find a proof currently in 3ds max... maybe I miss something but I don't see much stretching after subdividing them. I remember somebody said that using relax functions can cause problems when using n-gons... but I don't use it when I model with a sub-d surface technique.



In the end it all turns in quads, once you subdivide it. The question is how smooth the result is in terms of topology. We clearly know that we can't have a topology without E's and N's (a vertex shading 3 and 5 edges). Both a triangle and n-gon subdivided produce a pair of E and N, or a t-loop.

Using triangles and n-gons we reduce the amount of polygons as you will see further.

There are some places to avoid triangles and n-gons, but in some places it's quite usable. Places to avoid are joints and the face of the character, simply because we need a good and predictable deformation there.

So how i think they can be used? The point here is to remember where the pinching will be, i.e what kind of irregularity we get on the next level of subdivision.
The thing is, if we need a triangular structure, there will be a triangular type of subdivision anyway. The only difference I see is that if we use all-quads, we must add much more polygons, and if we use a triangle, we don't need to. So if using a triangle we get a lighter mesh in the end. But using all-quads give a a bit more control of the cage.
For example, if we take a look at folds at the forehead of the girl in the practical example, it's clear that only a direct t-loop (not a triangle or n-gon) is acceptable, as we need a fine control of each vertex to produce the required definition of the fold. But it's not always required.
Funnily enough, we get a t-loop from both a triangle subdivided and n-gone.


A t-loop examples.



What is clear that we get more control using n-gon if we need to make a volume like a muscle, so we just take that edge and elevate it. With a triangle we need to elevate 2 edges, but the difference is minor to me.
With n-gon we have only one controlling edge or elevation, so we can't make it more stiff as with triangle putting those 2 edges closer. So n-gon subdivision is a bit "milder" in a sense. We can simply turn it into a triangle by adding an edge though.



And now think, how much lighter the geometry is using triangle and n-gons for your profit, and how much edges we would have to add to make it all-quads.
The first man who I met using such a technique is Steven Stahlberg, and now I came to conclusion that it's damn good.

Last edited by mister3d : 10 October 2009 at 04:19 AM.
 
Old 10 October 2009   #37
@Jaco
So theres no confusion by me... By -triangular shaped quad-, topologicaly speaking, you mean this right?



That its pointed more evidently in Stepping has one of basic elements .

But, you are talking, not only of Opposed 3-5 Poles but any triagular shape with Opposed vertices with a difference o Two or more edge connections, like this, right?


So this kind of config offers bad calculations for cloth and so on? Ok. I didn't know that. Thank you for pointing out.
So for Stepping you would use only C-Loops (i call them a pair of trapeziums), like in the image?





@mister3D

I tryed some time ago something related to that topic.
It Indeed reduces the edges and in the feeze we can see less concentration.
Looking into animation deformation, the more edges you have, more work to control them, but seem that poles are easier to control.
I did not do more experiments on this :P



Originally Posted by mister3d: All-quads technique is predictable, ....

And still we have not yet seen (i think :P) a retopology tool using that to generate mesh automaticaly. Imagine you putting poles markers in a Volumetric Sculpter like in 3D coat and BAMM... Key Loops.

Cheers

Last edited by animatics : 10 October 2009 at 09:56 AM.
 
Old 10 October 2009   #38
Clothes is a special topic, but in 3ds max you usually use proxy - a simplified clothes representation with as uniform topology as possible.
I'm not sure what you mean by an automatic topology.. I think we will see topology obsolete faster than seeing an automatic one. Just look how the increase in gpu capabilities goes on. I predict in 5 years polygons will become obsolete for character modeling, and will stay only for a hard-surface one.

"Looking into animation deformation, the more edges you have, more work to control them, but seem that poles are easier to control." - same here, if you have too heavy geometry, you can use aproxy model to rig it. About "but seem that poles are easier to contro"l - bot sure what you mean.
 
Old 10 October 2009   #39
Ok, so i told you i would answear

Originally Posted by mister3d: -having more than 5-edged pole doesn't make sense, as you can make 6-edged into 3-edged, unless you make a wheel like in your example, which is rare for subdivision modeling

6-edged into 3 edged. How? i knwo how to make a 6 edges into two 5-edges, but not into 3.
I kept the max number of N-Edges to 5 in my last table.

Originally Posted by mister3d: -in your elbows example you should mention how many edges are needed to avoid texture stretching and why. The logics of elbow subdivision is pretty simple and could use less examples I think. Because all you need to know is how to make a rigid edge and a smooth edge

Yes i could have less examples, altough i think it's a matter of design. I was thinking in doing another one in wich one of the parts its an animated elbow.
Texture Stretching? I'm not following. Are you relating to the fact that i didn't used diagonal cut alone, for the last too, when i could have?


Originally Posted by mister3d: -in your extrusion of an 9-sided polygon there is a pinchiong problem, a very common in sub-d modeling. This is what would be interesting to hear how to fix

I have no extrusion of such polygon. I don't have a 9 sided in any of my images. By the way a 9-sided polygon is the one you want for a subpatch triangle. 3x3=9.

Originally Posted by mister3d:
-dividing a straight edge inbetween just to make it quad is not a good idea, because on the next level of subdivision you will have a pretty ugly pinching. Actually, it will be even more ugly than a triangle. The topology will look odd when you extrude the inserted vertex. So "internal point" is a bad idea imo, it's better to subdivide it with triangles then. If you are really about keeping it all quads you can subdivide it once and get your quads But this brings up the idea that subdividing a triangle is the same as subdividing a 3-sided pole, so having triangles may be not a big problem, at least in such extreme examples as yours, where they have one point elevated (so having controlled triangles is fine imo, but not everywhere without control). N-gon turns a 5-edged pole on the next level, by the way. Though keeping n-gons is ugly, definitely.

I'm not following this cut. Are you talking of creating a 3-Pole anywhere? what specific example?


Originally Posted by mister3d: -in your example with 12 and 36 vertices, wasn't it be easier to use hard-surface modeling? It looks really rigid, why to use it?

The idea was to show that multiplying by 3 is the way to calculate what type of surroundings you need to make a sharp edges in subpatch. so in the Table. That's what's stated in the Extrapolation of Central division row. so it will be like 3->9, 4->12, 5->15, 6->18, 7->21 8->24 8 (like the spehere with octogon) and so on. i just kept going for a matter of exemple.

Originally Posted by mister3d:
So I would really like to see:

-how to optimise an extrusion of a box without edges going all way long to the ends of the model
-fixing pinching problems


number 1, not following
number 2, me too =)

As for the insertation i'm not following what you are trying to acomplish. Is it the fact that the 5-edged are not in the edge contour?
_______________________________


So you think in 5 years topology is obsolete and ppl will model with n-gons all the way? hun... i don't know.
There will be always need for simple mesh for Dynamics calculations and i don't know how n-gons behave there.

The "automatic" retopo is something i'm thinking about, but i can't yet explain it right now. lol

By control, about the poles or n-gons, i mean that moving the pole around you know it will be there every subdivision level. With non-quads you have to move 3 or >4 points to control it's position. But i know it looks so smooth and working=) But still n-gons are still not allowed in some softweres or particular features right?

Cheers
 
Old 03 March 2010   #40
//testing!

I think a flash enviorment is the ideal for the Table since it can expand the contents further with concept explanation sections than the primary image withoutn the need of making a video.

I'm a flash newbie so this is my experiment for the Table using Flash, applyed on a remake of the Elbows sheet.

Any constructive critiques are welcome.

Cheers

-> Dowload
or
-> View Online

main page preview
Attached Files
File Type: zip Elbow_Models_lwo.zip (24.6 KB, 31 views)

Last edited by animatics : 03 March 2010 at 04:15 PM.
 
Old 03 March 2010   #41
Your host is currently checking your site for possible malicious content, unable to view it, but putting this info in a flash file seems it would work great.

I just stumbled onto this post, and i have to say it's a great ressources for beginner modelers (i wish i had something when i started school instead of all the trial and errors in topology that went on for such a long time to be able to know what i needed)
__________________
Itís the sensible, logical thing to do, of course, which is why we donít do it.
 
Old 03 March 2010   #42
thanks that's the feeling i was looking for.

Yes, you are not the first to report that host problem, anyway, it's avaible for download.

i think the thing has too many columns and could probably be kept simpler, but i don't know wich one to suprees. I could even discard the first row, since it might be the most complicated and useless...

Cheers
 
Old 05 May 2010   #43
Question

Long time no see mister3d
Originally Posted by mister3d: And now think, how much lighter the geometry is using triangle and n-gons for your profit, and how much edges we would have to add to make it all-quads.
The first man who I met using such a technique is Steven Stahlberg, and now I came to conclusion that it's damn good.


This was the part that made me relook to your post when i was looking for stuff and saw one of Steven Stahlberg's pictures and finally understood what you were talking about in this post of yours.

I think more practical examples, more evidences are needed in this area of using triangles and n-gons, especially of triangles opposed to the TriangleShapedQuad that Jacco and others said it was "evil", and that i recieve well, but my head asks "Why?".
(not that i'm going to do that right now, i'll just try to put some salt into the discussion)

I did these based on what i saw in you images (click to enlarge)


At first i thought they were decieving because you talk about being topological similar since they look almost the same after subdivision, but then you say the Ngon gives one control edge and the triangle, two. So i tryed to put them in an alike topological situation, but that didn't work very well so then i understood what you meant and try to put them into an even spacing situation.

Now, the N-gon it is lighter, but i had some issues with cloth and texture baking, wich can be one me and not on the N-gon itself. So in this line of thinking, i think i'll stick to the triangle.

-> Also i have a question for you. Using N-gon doesn't give issues when later you need to edit the mesh passing by the N-gon? I gave it a shot trying to reverse the detail in, and i wasn't very happy with it.

So i guess, quads and triangles make a better way and the N-gons, could be used just for final detail.

Cheers

Last edited by animatics : 05 May 2010 at 05:22 PM.
 
Old 05 May 2010   #44
Originally Posted by animatics:
Now, the N-gon it is lighter, but i had some issues with cloth and texture baking, wich can be one me and not on the N-gon itself. So in this line of thinking, i think i'll stick to the triangle.

-> Also i have a question for you. Using N-gon doesn't give issues when later you need to edit the mesh passing by the N-gon? I gave it a shot trying to reverse the detail in, and i wasn't very happy with it.

So i guess, quads and triangles make a better way and the N-gons, could be used just for final detail.

Cheers

If you push your edge too far up, you lose control over its texture coordinates, I'm not sure it has something to do with n-gons. You then can build polygons there and even detach the piece to get more control over the texture. Experiment with triangles and n-gons. Just use what you feel is right, without too much theoretizing. Some basic theory, and then dive in into modeling!
 
Old 05 May 2010   #45
thank you for your work !

.

 
reply share thread



Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
CGSociety
Society of Digital Artists
www.cgsociety.org

Powered by vBulletin
Copyright ©2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump
Miscellaneous

All times are GMT. The time now is 10:44 PM.


Powered by vBulletin
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.