Collada bone - skin separation in JME2

I just ported some of my programs from jME1.0 to jME2.0. But I find the collada model which is fine in jME1.0 is not working properly in 2.0.  The screen shot is as follow:



The skin works fine in 1.0:





Seperated with skin in 2.0.  The position of the skeleton is right, but the skin is facing to the wrong postion.





BTW, mulova has mentioned this problem earlier here: http://www.jmonkeyengine.com/jmeforum/index.php?topic=8652.msg67174

But I think maybe this is an internal problem of ColladaImport in jME2.0.



The model file is here:

http://www.yourfilehost.com/media.php?cat=other&file=ballMan3.DAE

For my experience with anim/skin in MD5 Reader 2, I can tell you that there is some object whose transformation is lost or erroneously concatenated (probably a Bone or the Mesh Object).



Wich software do you use for editing models and export?

Thanks for your reply!  I’m using 3ds Max 8 and ColladaMax 3.03 for model editing and exporting, and currently the skin/bone animation runs well under jME1.0.  There were some problems with the texture and animation of the outputed collada, for example,  only lambert and phong shading is supported in jME collada importer, so I apply textures in 3ds Max using this two methods.   And the outputed collada animation also has some problem, I have to edit the outputed dae file to make it work.  After solving these two main problems the collada format just works fine. 

I’ve tried to use MD5 in my game, but because of the buggy MAX/MAYA exporters for MD5, finally I gave up. 

And I find that there’s no means to control a single bone or a branch of bone of the whole skeleton (for example, by moving the mouse, the character moves up his right arm and performs an aiming gesture while the other parts of his body keep performing the animation loaded, which is very easy to archieved by using skin/bone system in jME) under the MD5Importer for jME1.0.

Furthoremore,  MD5 importers for jME have their own skin/bone system instead of reusing skin/bone system in jME.  This may put a risk of having too much dependencies on the MD5 importers.



So for the reasons above I finally desided to using collada in my game.



Here are some screen shots of my collada model test, take a look.  The jME collada importer works perfect!



Idle animation:





Idle animation bone debug:





Walk animation:





Walk animation bone debug:


I did not suggested you to use MD5. I only said that, because of my experience with MD5 Reader 2, I think that some object transformation is missed, during import (or during export).



You have to manage to modify jME COLLADA importer, if nobodyelse do it for you. I just suggested a possible direction where to search the bug.