PDA

View Full Version : topology issues encapsulating creativity!


lowkey
09-12-2003, 03:03 PM
hi all,

does anybody feel the same when modeling?

i always come to the point where thoughts about correct topology keep haunting my mind and holding me back from actually continuing with the modelling process. :thumbsdow

i know the whole topic is a science for itself and i also read a lot of research threads about it, but what am i to do to practically get my models done?! :shrug:

here's a simple example of such a situation:

lowkey
09-12-2003, 03:06 PM
in this picture the origin of doubt lies in that 5-sided vertex encircled in red.

how do i go on from there?!

lowkey
09-12-2003, 03:12 PM
do i split it up and get more 5-sided faces on my way, like to be seen in this image?!

lowkey
09-12-2003, 03:13 PM
or do i risk a triangle, which won't smooth correctly!?

lowkey
09-12-2003, 03:17 PM
things like that drive me crazy and result in deep frustation every time. :annoyed:

how to positition correct edgeloops, without creating too heavy geometry? how to efficiently avoid non-manifold geometry? tris or 5-sided!? etc. etc. ...

how do you work on your models?! do you model straight ahead and clean it all up later or do you plan every split before you actually start modelling?! :shrug:

please share your thoughts!

Jolts
09-12-2003, 05:40 PM
I would flow that line up into the eye or the cheek bone.

lowkey
09-12-2003, 05:55 PM
so i'll just have to take a look on hiding cumbersome geometry in unseen places?!

that's all? i thought there's some special technique on avoiding it on from the beginning! :hmm:

tredeger
09-12-2003, 06:04 PM
Well, this question certainly gets at why modeling *for animation* is sooo much harder than modeling for non-deforming objects. People like Bay Raitt (check out http://cube.phlatt.net/forums/spiraloid/index.php ) do amazing work unbelievably quickly and intuitively. Their edge looping method centered around box modeling is amazing and when done right seems to gaurantee proper (read quads) geometry. I can't seem to get my boxes to do this. Ever.

I'm working on a male human now that I've been revising forever as my ambitions toward a really realistic character have climbed. Some of the lessons I've learned:

1. You're placing your edge loops to capture the body's natural contours in a logical predictable and easily controlled manner. A lot of these contours are governed by musculature.

2. So, Learn as much as you possibly can about anatomy, physiology, and kineseology. Know where muscles are coming from and where they go is soooo vital.

3. Every edge loop is a trade off. The musculature shifts in incredibly complex ways. The entire shoulder girdle is an truly mind twisting example. If you look at the scapula slide around under the skin while the muscles attached to it change shape it seems impossible to capture all of the possible contours that could emerge through the range of motion. It's like trying to sculpt a mesh that can capture the surface of the ocean in motion. Just what is the shape of that wave?

3. Every edge loop is a trade off pt. 2 Certain areas of the body are more topologically complex than others. And they all run into each other. So it is incredibly difficult to capture adequate detail in one area without ending up with an overly dense mesh in another area. And the transitions between body parts is complicated by the way that muscles twist around eachother at the intersections. The shoulder is a great example (actually the shoulder epitomizes just about every hard modeling problem). Another example is the back of the knee. The biceps femoris and semitendenous muscles split to either side so when flexed, you get this depression in the crotch of the knee. Well, the calf doesn't continue on from the tendons of these muscles, but has its own origin just inside them (where the depression is). But when the leg is straight, this concavity becomes a convexity as the muscles push the bursa out toward the surface. And the size of this lump becomes a function of fatty tissue in the area. Sometimes visible, sometimes hidden, depending on position of the leg. And very difficult to capture topographically.

4. Every edge loop is a trade off pt. 3. We've already seen that most of your loops don't intersect eachother cleanly. (The ideal is a perpendicular intersection). But in addition, creating a flow in one direction for a particular purpose can run completely contrary to other needs. There is a good discussion of this over at the human head: http://coldfusion.art.msstate.edu/camenisch/thehumanhead/index.html ) Another example is the lateral portion of the torso. The topo lines should flow parallel to the ribs in front where they can be used to capture the form of the digitations of seratus anterior and either ribs or obliques depending on pose. It can be quite difficult getting a topo that achieves this without totally over loading surrounding areas with density. But these lines flow over to the back in a really problematic way. The lats and trapes are interrupted by the muscles of the scapula which themselves are quite chimeric in shape. But more to this particular point is this:
As the lines in front flow to the back you must decide whether they should follow the direction of the lats down toward the sacrum or go the opposite direction following the flow of the posterior part of the ribs. Either one can end up determining the dominant shape of the back depending on the pose. (The lats tend to be more prominant but not always, and the Langer lines follow the rib).

Is this driving you crazy yet?

So how has all this informed my technique? Well, I've taken to doing a lot of my planning outside the computer.

Pose a lot in front of the mirror. Watch things shift around and notice the transitions. Model for movement, not just shape.

Build physical models. 2D references are great but building a 3D rough can really help you understand your model. You don't have to get the proportions perfectly correct, concentrate on understanding and reflecting anatomical form. Toward this end, build the model up in layers. A simplified skeleton (spline and rib cage) then start attaching muscle layers. As you solve the problems of doing this in the right order and observe how the muscles intertwine you'll learn a great deal about your character.

I figure out my topology on paper first. I start with front, back and side view references then overlay sheets of artist's tracing paper (which is most convenient when purchased in rolls). On this I start drawing my contour lines and seeing where they intersect. I begin planning the flow of edge loops based on having to resolve these intersections. Then I start turning this basic outline into quads on paper, making sure that lines in one view flow over to the adjacent views. As the strips of quads emerge, the problems become evident even in 2d. The drawings look a lot like discontinous UV texture projections. And i find it is much faster to resolve these problems on paper than in the computer. Later I'll scan in my final topological solution and use it as a reference for the 3D version.

When planning your loops, try to get them flowing not just along the contours, but in a manner that has them terminate cleanly. If you can capture a broad area with a nice flow, then it makes it much easier to return to the area later and put in greater detail with an insertion of lines that are self contained. In other words, lines that don't create new topology beyond the local area. Toward this end...

You'll need to learn some general patterns that will greatly assist the problem solving that is required to create good topology. Learn different ways of turning a flow of 4 strips into 2 strips. How to split a 2 strip so that it forks and lets you insert new geometry at the fork. How to get not quite perpendicular lines to blend in x number of strips. Etc. Etc. I used to drive myself crazy by laying out mostly quads then ending up with a single many sided face that I had to try to chop up into quads and this sometimes proved impossible. I did learn to go for certain target shapes first--like a triangle next to a pentagon, which can always be recut to a quad. But this is still a devestatingly tedious undertaking. You can spend an evening trying to drive a single triangle across your geometry to some location where is can be properly resolved. And in so doing, you can spoil most of the nice flow of your quad strips. I find it far preferable to try to keep everything in quads straight through.

This all becomes a challange when you try to do your paper layout in parts. For instance, I spent a lot of time doing the sides/front and back of the leg and making sure they came together and had enough geometry to get all the muscles and this was all well and good untill it got up tothe top of the thigh. Then things start flowing strangely. So to continue, I need a prelimary set of lines for the abdominals (simplest part of the body as a series of stacked boxes), the obliques, glutes (simple on its own but HARD to integrate with the leg, obliques, and lats) and lats. And I know i don't just need equal numbers of lines coming out of the thigh as I have going into the torso. I need certain lines to match up, as they flow into each other. So I end up with important constraints to observe. My torso must either use all the lines feeding into it or cleanly, sensibly smooth them down or expand their number as needed. In fact, it has been the case for me that some areas need reduction while others require expansion. So in addition to tracing body contours, I'm always on the look out for areas where I can control flow *given* the number of lines I need for the surrounding areas. For example, you can be extremely flexible with the flow at the front of the pelvis (the crotch) as many solutions will satisfy the necessary shapes here. This is important as you must resolve the cross cutting of the sartorius against an otherwise clean anterior portion of the thigh.

Finally, I think it is helpful to think anatomically even when your characters aren't realistic in nature. You can simplify your thinking a lot, but the general principals can be really helpful even for cartoonish and fantastic designs.

Anyway, I'm running out of steam here. Hope this rant is somewhat useful. I'll try to post examples of what I'm talking about when I have more to show. Good luck with your stuff. As for the picture you showed, I'd either redesign your loops or try to drive thatpentagon deep into the oral cavity and hide it away where it will never be seen.

lowkey
09-13-2003, 08:31 AM
alright mr. tredeger, your reply is far off encouraging, but a very interesting read with lots of valuable information in it! :)

right now i'm short on time, but i'd like to continue this discussion by any means!

now, what i'm after doesn't reside too much on the artistic side of the process, but more on the technical. i mean, i already knew (theoretically!) that edgeloops help define the flow of musculature and anatomical structures, but what i'd really like to know is:

how do people approach problems like i posted practically!
(which is a very simplistic example)

i mean, if i have to have edgeloops e.g. around the mouth or eye area, how do i go about it!?

do i split it straight on up and fix the resulting 5 sideds and tris later on? do i have to spin my loops around the whole object to maintain continuity although i do not need that much detail in every area? how to get rid of that insecurity when building up meshes? do i have to experiment more or is there any form of 'in detail' literature available?

ambient-whisper
09-13-2003, 08:37 AM
basically. you want your model to be clean where most the deformation happens. so get rid of that triangle near the lips. and continue the loop until it runs into an area that doesnt deform much. once you hit that spot you can do whatever you want. as long as the subdivided model looks good.

thats what i do anyway.

kpalazov
09-13-2003, 08:35 PM
Absolutley,
You are very right to say this and I agree with you whole heartedly lowkey. I have had this quam for the past years now and as a result have changed my way of working. I would how ever like to restrict my view to organic modelling. Its the nature of industrial modelling to be more technical then intuitive .

Software today, as you have mentioned, is very restrictive in the respect that as you work to create some form and or feature your mind is involved in two way, creative process - bring the character to lfe- techical process- creating and fitting the toplogy. Simply put this work flow, unless you have alot of experience, is very inhibiting and more so a craft then the art form it should be.

Fortunantly for us there is a solution to this. But first consider a work flow which serperates these two tasks. That is creation of the form and features without the head ache of topology while modelling....and secondly the creation of the correct animatable topology over this form.

The solution comes in the form of ZBrush. I will not go into why and hows of the program but will simply say that the work flow I have mentioned above is entirly possible within zb. You may even consider creation of some very primitive inital volume in another application. An idea very similar to the work flow of creating a mesh skeleton in sculpture. Its really a pleasure to craate in zb as its so amazingly similar to modelling with clay, this enjoyed and ability to of self expression is by far enough motivation to help you conitinue and finish the character or whatever organic your modelling.
Zb takes care of form and feature...but what about topology.
Well I have been fortunate enough to have access to paraform. Its a tool used in converting point data to pologons as well as surfacing pologon data with nurbs . Using Paraform I am able to draw , literally, the topology over the high poly object I have generated in zbrush. Fitting exactly the curviture and features of the model as well as taking acount for any deformations .

And thats it basically a way of artistically creating characters and organic form quickly and eaily, with this way of working you are basically free to create characters to the fullest of your understanding of anatomy, form and features. Despite the fact that Paraform is not widely available to most people, due to costs, I'm sure that other programs will follow suite. For me this is the next step forward in computer aided modelling .

I have not mentioned alot of details and other advantages like being able to take the re-topologies surface, generate displacement and normal maps and render a high poly version of your charcter while animation with the lowpoly...plus many other but I have you have some perspective now on the things I have mentioned.

All the best and good luck

Kiril

Marcel
09-14-2003, 01:28 PM
I think you worry too much about the technical aspects too early. Some things you can only learn by experimenting and by building up your modelling experience. Like the shoe brand says: "Just do it!" :)

A thing that might help you when you feel that you are stuck is to look at the topology of other peoples models. Whenever I see a render of a great model I always save the render and the wireframe. It can be very helpfull to see how other people solved certain problems.

Peter Reynolds
09-14-2003, 02:16 PM
I think your biggest issue is FEAR of making a mistake. This is causing you to stall during the modelling process.

You already know the importance of edgeloops. You already know they should flow in accordance with muscles, etc.

Often there is more than one solution that will work, that will describe the form and and muscle flow and keep everything quads.

But if you let yourself make mistakes, you'll learn a lot faster. You'll probably learn more from doing 5 quick faces in a day, than worrying about mistakes and then trying to produce the perfect solution over a period of 5 months.

But when you're really in doubt, just look at some of Bay Raitt's wires and the solution will often present itself to you.

NO FEAR!

lowkey
09-14-2003, 02:54 PM
hell yeah, thanks a lot guys, i know it's a very deep topic to cover and i'm glad to find that many helpful thoughts in here.

You'll probably learn more from doing 5 quick faces in a day, than worrying about mistakes and then trying to produce the perfect solution over a period of 5 months.

that's so damn true! i really don't know what i'm afraid of?! instead of 'just doin'' it and evolving during the learning process, i'm rather stickin' my head in the sand, worrying about theories! :hmm:

resume:

now, what seems to be most important is keeping the deformable areas as clean as possible.

so does it really matter if non-animated areas contain tris or 5-sideds as long as they smooth OK?! :shrug:

after having read all the info in here, i'd guess it doesn't, which encourages me a lot!

well, i think my main problem was the wish to get it all cleared up in the head (theoretically), before simply starting it off practically.

that's what held me back all the time.

i mean, i scrutinized other peoples' wires a lot, but never while i was actually working on my own meshes.

it sounds weird, but that's how it is! :D

keep it coming folks, i like to get more reports on how people approach their geometry!

Marcel
09-14-2003, 04:21 PM
Haha, this thread it's sounding like a self help group with all the talk about fear, but it is very true! :) You need to experiment, I learned a lot of things from 'happy accidents'.

Here is an example of how I start a mesh. It's a caricature head, so the proportions are very weird.
As you can see it is a complete mess, there are a lot of unterminated edges. The only thing I am trying to do now is to find the topology that enables me to make the correct shapes (the nose and chin are ok right now, I'm working on the cheek at present). Once I have that topology I will worry more about how to smooth out the surface and how to get rid of the tri's and five-sided polygons (if nessecary).

If you look at the wireframe cages of good modellers you'll see that the faces/edges are following the shapes of the face. Try to find out how they do that because it is very important if you want to make a good model.

lowkey
09-14-2003, 06:03 PM
ehhehh, yeeeh...reminds me of that scene in 'fight club'! :D

but it really helps a lot!

Peter Reynolds
09-14-2003, 06:19 PM
you do NOT talk about topology club!

zef
09-14-2003, 11:27 PM
I have been suffering from the same symptons for these past few weeks. :blush:

I'm still a n00b to 3D so I have alot to learn. I was about to give up on it all together, but just kept pressing forward. I really appreciate the responses in this thread; they helped with alot! :beer:

Thanks alot guys! :buttrock:

lowkey
09-15-2003, 07:59 AM
you do NOT talk about topology club!

muuuuahahHHAHhahhHAhhHAHAAAWW!
sorry, about that!:D

Neox
09-15-2003, 09:19 AM
don't care too much about it, you'll see that all tris and ngons will start to disappear, its magic ;)

lowkey
09-15-2003, 09:28 AM
you'll see that all tris and ngons will start to disappear, its magic

you mean, they'll disapper with growing experience!? i definitey hope so for i drove myself nuts with it long enough! :thumbsup:

Neox
09-15-2003, 09:49 AM
ok jetz auf deutsch :p
Reg dich nicht zu sehr drüber auf, das bringt dir nix, model einfach drauf los! du wirst sehen, dass du mit der Zeit automatisch dafür sorgen wirst dass keine Dreiecke oder Fünfecke in deinem Mesh auftauchen und wenn dann verschwinden sie ganz schnell wieder, kümmer dich lokal nicht um sie, global wirst du feststellen dass ein Loop aus dem Auge auf einmal ne Super Möglichkeit im ein Dreieck auf der Wange wegzubekommen. Ähnlich passiert das mir andauernd, ich kümmer mich garnicht mehr um Dreiecke und Fünfecke und meine jetzigen Modelle haben wenn dann nur gaaanz wenige dieser unschönen Meshfehler und wenn dann nur in so stellen wie dem Nasenloch oder im Ohr...

sorry guys had to write that in german ;)

lowkey
09-15-2003, 10:05 AM
cool, bin echt beruhigt zu hören, dass auch andere sich mit diesen problemen 'rumplagen und trotzdem erfolgreich modellieren, ohne sich in theoretischen sophistereien völlig zu verlieren. :)

'just do it & fix it later' scheint mir ertsmal ein guter ansatz zu sein, die künstlerische seite nicht vollends aus den augen zu verlieren!

ich mein', wenn man das posting von tregeder liest wird einem ja angst und bange, auch wenn's nicht unwichtig ist...:eek:

hab' übrigens eure letzte international challenge verfolgt und muss sagen:

hut ab! :thumbsup:

also danke für euer feedback leute!

-
translation:

i'm glad to hear that other people have the same problems and still model successfully without losing themselves in theoretical pitfalls!

'just do it & fix it later' seems to be a good attitude without loosing the artistical side of the process.

i mean, reading tregeders post made me really anxious about it all, though it contained points of importance...

neox, i followed your last international challenge and must admit:

you guys rock!

so, thanks for your feedback people!

Squeakypics
09-15-2003, 12:42 PM
What a relief!
So it isn't just me that gets topology overload.

I've recently started playing with polygon modeling after getting completely exhausted with splines in Animation:Master.

My experimental workflow at the moment is to rough out the basic low poly shape using ZSperes in ZBrush (FUN!!!!!!) and then importing into SILO to 'Cut some Loops'

Well thats the plan. I haven't actually got round to the cutting bit yet (Chicken!)

Phase 2 will be to paint the object in ZBrush (FUN!!!!) and maybe rig and animate in Messiah. (Or even LW8 at this rate)

Good thread though!

Marcel
09-15-2003, 09:33 PM
I'd like to a a small thing to the discussion...this whole 'just do it' philosophy is not only for adding stuff, but also for taking away vertices/edges/polygons.
It can be very scary to delete parts of your model you have so painstakingly built, but that is the best option in some situations.
(Just because you have put a lot of time into it doesn't mean it should never be changed anymore.)
Sometimes if you are stuck it helps to delete a small part an start over from scratch on that certain piece.

In the attachment you'll see the cheek part of my model getting very dense. While more polygons means more control, there is a limit to how many polygons is still workable. If you add too much edges you will have to move too much vertices to make a change in the shape: your tweaks will take much longer. Also, with less polygons your subdivided mesh looks 'cleaner' (it is much easier to make it nice and smooth, which is nice in this case).

That's why I decided that one of those edgeloops just had to go. Luckily for me the edgeloop ran from the eye to the corner of the mouth in a rather straightforward way.

In other situations I might have to do some work to clean the geometry up, but as Neox said before: "the geometry will fix itself, tri's will dissapear like magic!" :)

lowkey
09-16-2003, 07:59 AM
It can be very scary to delete parts of your model you have so painstakingly built, but that is the best option in some situations.

yes, that's what i've also experienced a lot...getting rid of geometry often helps in understanding what's really needed and what's not and how to build things up in a more suitable way.

though it might feel odd to actually do so! :eek:

Neox
09-16-2003, 08:13 AM
you can train that pretty simple with doing more lowpoly works ;)

CGTalk Moderation
01-16-2006, 02:00 AM
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.