Polygons vs. Triangles

[KeeperS Karu]

Heyla and well met!

 

I've been trying to learn as much about uploading meshes as possible. LL's definition of mesh is as follows:

"Mesh - A collection of triangles with a single transformation matrix, roughly equivalent to a "Prim" in SL".

This definition made me pause, because when playing around with meshes, over the years, the faces on the mesh were always referred to as polygons in the modelling software I've played around with. More, the polygons in the mesh don't necessarily begin as triangles. Some may end up that way, but not all of them. It just depends on how things turn out while I'm shaping things. Does the fact that the polygons aren't actually triangles matter when uploading them to SL? Or will I need to pop the mesh into yet another program that will convert the polygons into the acceptable format? Or, perhaps, is this something that is automatically accomplished when the mesh is exported as a COLLADA file?

For reference, I have found that I actually enjoy using Anim8or to create my models. I've tried both Wings3D and Blender, but the UI in either programs were simply not user-friendly for beginners. When creating my models in Anim8or, I would then export the models in .OBL format, UVMap the model, create groups and material regions, import them into Poser, rig them, if necessary, then use them, as needed, in my Poser scenes. After some reading, my process seems to remain the same, with the only additional step being that, when my project is complete, I will need to export the scene in COLLADA format, which Poser Pro 2010 seems to support (I'll be doing some experimentation to see how well this actually works, in practice). However, after reading about the reference to triangles, rather than polygons, I'm wondering if I need additional steps in order for my meshes to work within SL?

 

Thank you for your help!

Keeper S. Karu

[Marielle Caerndow]

Many 3D-creation tools can display and work with polygons (3-, 4- or more-sided).

However, internally these n-sided faces are (internally) made of multiple triangles. In example a 4-sided polygon ("quad") consists of 2 triangles.

It's can be important to not forget that on "base level" models nearly always consist of triangles, even when the program displays quads (or n-gons).

 Important because:

1. Game engines (SL) mostly work with triangulated meshes. If you upload a mesh to SL, it's faces are converted to triangles...
2. For the performance (GPU), the triangle count matters, not the 'poly-count'. If you have a model made of 100 quads (4 sided polys) , then it's triangle-count is actually 200. 
3. When animating a mesh, differences can occur. If you animate a quad (or n-gon)-mesh in the modelling-app, it can look a bit different in SL (since the mesh got triangulated on export/import).

As general rule, I model as I like it - with triangles and polygons. But before exporting the model, I convert all faces to triangles (since this happens anyway on export). So I can spot possible issues and get the 'real' triangle-count.

[Gaia Clary]

Maybe the Step II in my Shoe_quest can give some more hints. There i touch this issue very briefly. You need to scroll down a bit to see it.

 

[KeeperS Karu]

Ah! Thank you, both, for clarifying this issue This helps, greatly!

[KeeperS Karu]

I would very much love to see you put out a separate tutorial on the secrets of tris and quads! I can't be the only person asking questions concerning them... I hope. :matte-motes-bashful-cute-2:

[Rage Riptide]

I beleive the best way is to model in quads only, if possible, with a nice, neat topology flow. Some say triangles are fine, but I always find they get in the way at some point. Triangles become a problem when you want to add or remove an edgeloop but one must keep in mind that the mesh will be converted to triangles at completion.

So, convert to triangles as the last step before export and examine the results, usually fine and dandy if the topology is nice to begin with. If there is a problem then it will be much easier to correct.

[KeeperS Karu]

---

Rage Riptide wrote:

I beleive the best way is to model in quads only, if possible, with a nice, neat topology flow. Some say triangles are fine, but I always find they get in the way at some point. Triangles become a problem when you want to add or remove an edgeloop but one must keep in mind that the mesh will be converted to triangles at completion.

So,

convert to triangles as the last step

before export and examine the results, usually fine and dandy if the topology is nice to begin with. If there is a problem then it will be much easier to correct.

---

Ah, so keep the model, as is, and if anything looks weird, it can be fixed while still in a state that's still easy to edit?

[WolfBaginski Bearsfoot]

That's about it.

 There will be situations where a triangle is needed to get the modelling looking good. A quad can be triangulated in two different ways, and sometimes it matters which you use. So don't be afraid of putting triangles in to get the shape right.

[Gaia Clary]

as you say ;-)

 

[KeeperS Karu]

Ah! Thanks for  the extra clarification! That really helps!

[Feynt Mistral]

Polygons (or polys for short) versus triangles (or tris) depends largely on your background as a modeler.  To the world of cinematography and art in general (those focused on still renders rather than animations) a polygon represents any number of sides from 3 onward.  In gaming and similar real time production fields, modelers will tend to use polys and tris and synonyms.  When I talk about the poly count of my meshes I refer to the triangle count (because triangles are polygons).

Many people swear by doing quad modeling.  The goal is to keep the edge count down so that you can clearly see how your edges flow.  This becomes important with character models mostly, as it lets you clearly deliniate muscle groups.  Triangles tend to limit geometry or require extra cutting to eliminate, however they're valid tools when you really just need to condense your geometry.  And as mentioned, in games (and all real time simulations really) your mesh will be converted to triangle strips for rendering.  It's much faster (for graphics cards) to work with an array of vertices, with each current plus the previous two vertices being used to make a triangle, instead of quads where there isn't necessarily a clear way to create a strip in an array form.