Problems with and questions about proxy mesh creation

[crashoverride]

Some questions about proxy meshes and defineProxy.py http://sites.google.com/site/makehumandocs/proxy-meshes

- Which revision of defineProxy.py was used to achieve working proxy model (I don't get working one with the current script)?
- Which version of blender was used (graphicall or official)?
- Which daily build was used to export "default mesh" (the .obj that is used as Human model for proxy making)?
- Which python version was used?
- What operating system was used?
- These questions are for those who have already made through the proxy model creation work flow (you have working setup)

These questions because:
- I really would like to make my own proxy mesh since I've got dozens of character models to make for our game (wouldn't like to make the game model cleaning by hand or at least would be faster to have good proxy model )
- I could also give the proxy for you everyone to use (I like to make cleaned and optimised version but not too low poly)
- I have banged my head against the wall for about 13 hours (with co-worker) in attempt of getting this thing to work
- Got problems with face to vertex association (and yes I've tried with triangulated and non-triangulated model, didn't work)

I can provide more detailed debugging info if this info isn't enough .

And thank you for the answers

[ThomasL]

- defineProxy.py in current svn

- Blender from graphicall. The proxies were created with versions from a few weeks ago, but I tried the most recent one today without problems.

- Python bundled with Blender. I use builds by Fish on Ubuntu, various builders on Windows.

- OS Ubuntu 10 64 bit. I usually check mhx export on Windows XP 32 bit at work, but am unsure if I ever tried to build a proxy file there.

- The version of the exported default mesh does not really matter. Recent versions (from a few months back) differ at the eyelashes or on new joint diamonds, but you probably don't care about those. These additions have higher vertex numbers than the interesting parts of the mesh, and thus does not affect it.

For reference, I have uploaded the blend that I used to create ascottk's proxy file: http://rapidshare.com/files/399515615/ascottk7.blend.rar It works with an graphicall build from yesterday

There has been some changes to how a .proxy file should look. The first line now specifies the name. Also, the Faces and TexVerts sections may be replaced by an ObjData section, cf. ascottk's mesh. The .proxy file should look either like:

Name name_of_the_proxy
Verts
Barycentric coordinates for vert 1
....
TexVerts
Texvert 1 coordinates
....
Faces
Face 1 vert numbers or
Face 1 vert/texvert numbers
....

or

Name name_of_the_proxy
Verts
Barycentric coordinates for vert 1
....
ObjData
Content of obj file, with vert coordinates removed. It is a good idea to remove lines that have to do with materials.
...

Here are some things to think about:

1. Every vert in the proxy mesh needs to be assigned a vertex group (additional vertex groups are ignored).

2. Vertex groups with the same names must exist in the hi-poly mesh. Unassigned verts are ignored. In particular, don't assign vertex groups to joint diamonds, eyelashes, eyebrows, etc.

3. Apply loc/rot/scale to both meshes.

4. The script is not so polished. You need to change the output path at the end of the script (unless you have a /home/thomas folder).

It would be helpful to see the error output. You don't need to copy the entire output, just the final error and a few lines above it.

[crashoverride]

Thanks for the answer, I'll see if I get results with your file and get back to you in any case .

I have no problem running the script but rather with the end result (it becomes messed up ). The faces seem to be all wrong (totally). I think it has something to do with Barycentric coordinates for verts (mainly they are not what they supposed to be, well at least when I tried with Rorkimaru... tried to dublicate the results ).

More info again when I get testing (just woke up ).

[crashoverride]

Ok, I may have resolved this problem

What I did to the base mesh before was remove the diamond shaped objects which seem to be crucial . Will continue testing.

[ThomasL]

Yes, removing the diamonds will affect interesting verts with higher vertex numbers. It would be better if verts were ordered so that the diamonds come last, but this is not the way it is.

[crashoverride]

Here's the final product (if anyone is interested). It's quite high poly but still it only contains half of the original and has separate fingers and toes as well as inner mouth and teeth.
Vertices: 6822
Faces (polygons - mainly quads): 6454
Faces if trissed: 12628

Tested several times: Only had to do some correction with lips and fingers+toes (only if nails are important). Lips UV is not optimal if normals are baked .