PDA

View Full Version : A solid base topology


LetterRip
12-20-2004, 08:25 PM
Hi,

We've recently introduced a new mesh for the 'MakeHuman' project. To learn more about the Makehuman project you can visit this website,

http://www.dedalo-3d.com/

This new mesh is intended to address some of the issues raised regarding the previous Makehuman mesh, especially regards topology and such aspects as the ability of the mesh to subdivide smoothly, to have edge loops that support clean animation, a topology that can be easily reduced so that the mesh is useful for games and other low poly applications, and most importantly, a topology for which morphs can be easily created to change aspects of appearance such as gender, age, ethnicity, and individual aspects of character, as well as have the flexibility to provide for a wide range of expression and emotion.

We believe that the new mesh is largely up to the task, but we would greatly appreciate feedback from the thoughtful individuals who frequent this forum.

Our first version of this mesh was created by Kaushik Pal, with some recent changes by Manuel Bastioni to accommodate ideas gleaned from examining the postings and mesh on facial topology, especially those of Stahlberg

Stahlberg
http://www.androidblues.com/topology.jpg

and Laa-Yosh.
http://www.bunk.cistron.nl/plaatjes/poly_regions.jpg


Also for the hands Manuel has partially incorporated the excellent tendon and knuckle ideas of Laa-Yosh as expressed in this thread,

http://cube.phlatt.net/forums/spiraloid/viewboard.php?BoardID=2
http://maxrovat.sns.hu/nm/hands1.jpg

Now we would like seek your feedback on the rest of the mesh, to perhaps determine if there are any problem areas that we should turn our attention to.

First here are overview shots showing the default mesh with two levels of smoothing - We have just under 10,000 faces and 10,000 vertices, compared to a typical poser model which has about 30,000 faces and vertices.

http://www.dedalo-3d.com/lab/front_final.jpg
http://www.dedalo-3d.com/lab/rotated_front_final.jpg
http://www.dedalo-3d.com/lab/side_final.jpg
http://www.dedalo-3d.com/lab/rotated_back_final.jpg
http://www.dedalo-3d.com/lab/back_final.jpg

Next is the head and neck. As noted above, we've attempted to incorporate many of the ideas from Laa Yosh and Stahlberg, This image shows the incorporation of the major loops into the face mesh

http://www.dedalo-3d.com/images/new_head2.jpg

and here we see the incorporation of the prominent creases and wrinkles

http://www.dedalo-3d.com/images/new_head3.jpg
http://www.dedalo-3d.com/images/new_head3.jpg

now we turn our attention to the neck, here we see the flow of the sternocleidomastoid

http://www.dedalo-3d.com/images/old-new15.jpg

and then the adams apple and the point where the sternum meets the clavicles.

http://www.dedalo-3d.com/images/old-new16.jpg

LetterRip
12-21-2004, 01:34 AM
Next we turn to the upper torso and arms

http://www.dedalo-3d.com/images/old-new13.jpg
http://www.dedalo-3d.com/images/old-new18.jpg
http://www.dedalo-3d.com/images/old-new14.jpg

then the shoulders and back

http://www.dedalo-3d.com/images/old-new9.jpg
http://www.dedalo-3d.com/images/old-new10.jpg

and lastly the lower extremities including the gluteus

http://www.dedalo-3d.com/images/old-new2.jpg
http://www.dedalo-3d.com/images/old-new1.jpg

and the legs

http://www.dedalo-3d.com/images/old-new7.jpg
http://www.dedalo-3d.com/images/old-new4.jpg

You can either post critiques here or email me. Or, for those that would prefer to examine or work on the mesh directly I can email you the mesh in either a Blender .blend format or obj file format.

Thank you for your time and consideration,

Tom M.

pollywoggles
12-21-2004, 02:15 AM
Wow. Incredible. Excellent work.

harkyman
12-21-2004, 02:36 AM
Hey Tom - is there a Blender project that you're not involved in? Anyway, I was looking at this mesh a couple of weeks ago on the MakeHuman site. It's really an incredible step. I was interested in your initial comments, though, about the mesh being easily reducible for lower-poly work. Was that just a tossed-out comment, or is there some kind of methodology involved in that? Is the MakeHuman project considering using both a hi and low res set?

The reason I'm asking is that BlenderPeople will be best used with characters of varying resolutions to help with rendertimes and memory management. I was hoping to use MakeHuman in the more final stages, but level of detail was one of my concerns.

LetterRip
12-21-2004, 03:00 AM
pollywoggles,

I'll pass your compliments on to the modelers.

Harkyman,

I was interested in your initial comments, though, about the mesh being easily reducible for lower-poly work. Was that just a tossed-out comment, or is there some kind of methodology involved in that? Is the MakeHuman project considering using both a hi and low res set? The original makehuman mesh decimated very poorly. I've been told that this one does so much better. I believe there are plans for a low res version of the mesh as well though. I've also been looking at multiresolution mesh/mesh density reduction algorithms that retain topology (there are some interesting methods that achieve quite a bit better results than blenders current implementation - I seem to recall mention of a python script that does a better job, also there is a simple multiresolution implementation in the cal3d exporter...). So hopefully we will be able to apply the morph targets, and then do a resolution reduction step. (probably will require a few bits of python that are makehuman specific for the belly button and other areas...).

LetterRip

dienben
12-21-2004, 08:42 AM
Hello,

Wow, amazing workhttp://cgtalk.com/images/smilies/clap.gif

I learn a lot by this post, thanks for your work and your altruism.

Dienben

Bmanuel
12-21-2004, 09:48 AM
Hello,

Wow, amazing workhttp://cgtalk.com/images/smilies/clap.gif

I learn a lot by this post, thanks for your work and your altruism.

Dienben
Thanks!
To know the details about MakeHuman project, please visit
http://www.dedalo-3d.com/

CDI3D
12-21-2004, 11:21 AM
As a long time user of Poser and supporter of DAZ3D's figures I am following the MH project as closely as I can and with no small amount of excitement. This is, IMO how DAZ should have proceeded.
I can easily see MH becoming a rival quickly with the addition of this new mesh.


Tom: do you have the capability to show the mesh in its default androgynus state? And as a request to pass along to the modelers, could the defualt andro state not have female genitals by default, just a smooth crotch area?

Giangmatric
12-21-2004, 01:26 PM
Woh:applause:http://cgtalk.com/images/smilies/thumbsup.gif Good work!

llunasol
12-21-2004, 02:49 PM
Just to add some feedback :) The new mesh look very accurate. Looking at the differences we can see how horrible was the old basic mesh. The "whole" MH was an impressive piece of software but tweeking the mesh always produced sad results and I finnished to see it more a curiosity than a tool, but the new mesh puts everything on the right path. Is there any date for the next alpha release with the new "usable" mesh?

Many thanks.

MSB
12-21-2004, 04:08 PM
the most stuff I like it the foreArm improvement I used to add Lattice and SDK or Bla bla bla..

and the upper leg topology is great

But the knee -Back area need some work I think if u notice how its look I think u need some modification...I think the old mesh is better

the neck is great...

the breast is little weird..I think they are little "streatch"...


great work

Bmanuel
12-21-2004, 08:36 PM
But the knee -Back area need some work I think if u notice how its look I think u need some modification...I think the old mesh is better


Hi!
Note the muscle insertion can be easily obtained moving some verts, so I'm not sure the old mesh was better.

http://www.dedalo-3d.com/lab/k-mesh25.jpg



http://www.dedalo-3d.com/lab/k-mesh24.jpg


I think the old mesh is untidy... have you some example?
Ciao!

joo
12-22-2004, 02:26 AM
Thanks for sharing!!:thumbsup:

MSB
12-22-2004, 03:59 AM
Hi!
Note the muscle insertion can be easily obtained moving some verts, so I'm not sure the old mesh was better.
I think the old mesh is untidy... have you some example?
Ciao!

sorry no examples but I look back to my anatomy books and my sister books "she is doctor" to make sure.
The "Gastrocenmius" under the skin looks like two piceces but from outside looks like one piece.I think u should look to anatomy books.


The end of "Tibia" is not correct I think u should make it rounder.
The end of "Fibula" also.

cheer.

Bmanuel
12-22-2004, 08:23 AM
sorry no examples but I look back to my anatomy books and my sister books "she is doctor" to make sure.
The "Gastrocenmius" under the skin looks like two piceces but from outside looks like one piece.I think u should look to anatomy books.


The end of "Tibia" is not correct I think u should make it rounder.
The end of "Fibula" also.

cheer.
I agree this is not perfect, but sure it's better the old one, because we must think not only to a realistic modellation but (very important) to a easy rigging too. The old mesh, with various diagonal edges and even triangles, was very hard to bend...

I'm searching some 3d modellations from gurus.
thx!

Apollux
12-22-2004, 04:07 PM
Jaw dropping !

Is this the same mesh that is sopposed to come pre-rigged, or that is still for another release?


From now on I'll have a really hard time justifying myself not using MH... http://cgtalk.com/images/smilies/thumbsup.gif

Bmanuel
12-22-2004, 09:47 PM
sorry no examples
I've seen this, from Stahlberg: it seem similar to the K-mesh, or at least very different from old makehuman mesh:
http://www.androidblues.com/body5screen5.jpg



Apollux: the new mesh will be in the next MH release.

peanut
12-25-2004, 02:43 AM
Its good its good

but there's always a few places that need some work specialy edgeloops that look funky to my screen. Try to have a good flow of edge, its not important what topo you use watchout for cuts that defy the purpose.


http://www.dedalo-3d.com/lab/k-mesh25.jpg

ambient-whisper
12-27-2004, 10:25 AM
it all depends what you want to do with the model in the end.

the best setup that ive seen so far that is very flexible, clean, and great for zbrushing later looks like this.

http://209.132.69.82/uploaded_from_zbc/200407/user_image-1089870557tkd.gif

because it can later be turned into the picture below with ease. :)

http://209.132.69.82/uploaded_from_zbc/200407/user_image-1089746374ezr.jpg


its super easy to model too, because you start with a simple cylinders

Laa-Yosh
12-27-2004, 04:04 PM
Hey guys,
First of all it's so cool to see that people find my wireframes useful :)

However, I'd like to warn you that the stuff there might not be that good. It's because I've modeled those characters before I've had actual experience with scuplting facial blendshapes for animation - and thus, there are several things that I would do very differently nowadays. Which brings me to my other point - you should test the topology in action as much as you can before declaring it as final, especially for this project.

For example, on the mesh in this thread, I would certainly rework the cheek and jaw areas, because both the edgeflow and density are not fit to get some nice skin sliding there. And maybe also take a look at the whole eyebrow area...

Bmanuel
12-27-2004, 04:26 PM
Hey guys,
First of all it's so cool to see that people find my wireframes useful :)

However, I'd like to warn you that the stuff there might not be that good. It's because I've modeled those characters before I've had actual experience with scuplting facial blendshapes for animation - and thus, there are several things that I would do very differently nowadays. Which brings me to my other point - you should test the topology in action as much as you can before declaring it as final, especially for this project.

For example, on the mesh in this thread, I would certainly rework the cheek and jaw areas, because both the edgeflow and density are not fit to get some nice skin sliding there. And maybe also take a look at the whole eyebrow area...
Hi!
Thanks for suggestions.
Do you mean simply increase the density of mesh, or you think the edge loops is wrong?
Thanks again, and sorry for my poor english.

Bmanuel
12-28-2004, 12:37 AM
Modified head:

http://www.dedalo-3d.com/lab/new_head10.jpg

http://www.dedalo-3d.com/lab/new_head11.jpg

VM
12-28-2004, 04:50 AM
I agree with Tamas on that each face should have it's unique topology, dictated by different and specific needs. Of course, having a base model to start with is very appealing... :) and I guess you can draw many topologies on that base.
But I think this "face specific" issue is a big one. I see a world of clones... hi hi hi, where all loops are the same, and especially after being smoothed, the results are the same - like that guy working on a gun factory and stealing parts, to assemble them later, and he goes: Man, I don't get it, no matter how I put them together, it still doesn't look like a motorcycle...

Bmanuel
12-28-2004, 07:49 AM
I agree with Tamas on that each face should have it's unique topology, dictated by different and specific needs. Of course, having a base model to start with is very appealing... :) and I guess you can draw many topologies on that base.
But I think this "face specific" issue is a big one. I see a world of clones... hi hi hi, where all loops are the same, and especially after being smoothed, the results are the same - like that guy working on a gun factory and stealing parts, to assemble them later, and he goes: Man, I don't get it, no matter how I put them together, it still doesn't look like a motorcycle... OK, the topology above is wrong, I'm redoing it.
I agree with you that each face is different, but not so different. I agree totally with Mr. Stahlberg, here: http://www.cgtalk.com/showthread.php?p=354679#post354679

IMO it's bad to reinvent the wheel every time you need a new face, good point about needing to keep practising but there are other ways to do that.
Humans are genetically similar to 99.97 percent or something like that. Yes the lines in the face look different, but we all have the same underlying topology, the potential for those lines. It's like lines in the hands of our palms, all slightly different but basically very similar - for instance you'll never see someone with the lines going in the opposite direction, from the thumb to the pinkie.
Besides, consider the power of new displacement mapping and normal mapping...

LetterRip
12-28-2004, 08:28 AM
if you examine such programs as FaceGen, a topology of about 900 faces can handle nearly every natural human facial topology (minus wrinkles).

LetterRip

Bmanuel
12-29-2004, 10:28 PM
Update:

http://www.dedalo-3d.com/lab/new_head20.jpg

http://www.dedalo-3d.com/lab/new_head21.jpg

http://www.dedalo-3d.com/lab/new_head13.jpg

http://www.dedalo-3d.com/lab/new_head22.jpg (http://www.dedalo-3d.com/lab/new_head22.jpg)

http://www.dedalo-3d.com/lab/new_head23.jpg

http://www.dedalo-3d.com/lab/new_head25.jpg

http://www.dedalo-3d.com/lab/new_head26.jpg

VM
01-02-2005, 11:13 PM
I think topological models of ideal humans should be only the basis (wait, wait, I'm explaining), the beginning of the modeling process, and that the more detail you add, the more unique the end result.
What I want to say is that starting with a detailed base-mesh, a topology that pretty much covers all... it's risky. It's a bit like using a super-advanced rig, so advanced that the animator has to fight its features, because the rig is too automated, and not really flexible. In the case of modeling, that would be the defaultish head. I love the idea of the 'ideal topology', but I think it should be used at a fairly high level of modeling - when you only have a few polygons and the shape is raw.

Bmanuel
01-03-2005, 08:12 AM
I think topological models of ideal humans should be only the basis (wait, wait, I'm explaining), the beginning of the modeling process, and that the more detail you add, the more unique the end result.
What I want to say is that starting with a detailed base-mesh, a topology that pretty much covers all... it's risky.
Yes, I agree. It's risky...and hard. We should work with prudence, trying to find the equilibrium between details and general topology. For these reasons we post here step by step the progress of work...
We should have the 'common factor' between the most used human mesh topology.

Bmanuel
01-23-2005, 09:39 AM
The model will be available for public download soon, under Creative Common license.
In the preamble we write:

""
LICENSE

Thanks to a long study on tutorials and professional works (as the free models by Kaushik Pal)
and thanks to suggestion from great 3D guru as Steven Stalberg and Tamas Varga from CGtalk,
Manuel, Cicca and Z0 have modeled a new base model.
It's released under the terms of Creative Common License (Common Deed).
etc...etc...

"""

If we can't use the names above, please tell me, and I remove all as soon possible.
Thanks again,

Manuel

CGTalk Moderation
01-23-2006, 10:00 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.