mocap tool / animation retargeting

[ppClarity]

I saw this on the blog:
http://makehuman.blogspot.com/2010/09/f ... -tool.html

Like Thomas mentions, rig retargetting is complicated. I've been looking for explanations how to do it myself. Everytime I google the topic I keep running into the Hecker (Spore) paper. I keep meaning to sit down and bull through it but I keep failing my saves vs. confusion and sleep Is there anything a bit clearer or less complicated? A brief skim of the Poirier looks promising.

Here's what I've found today:

Code: Select all

@inproceedings{sporeanim,
 author = {Chris Hecker and Bernd Raabe and Ryan W. Enslow and
           John DeWeese and Jordan Maynard and Kees van Prooijen},
 title = {Real-time Motion Retargeting to Highly Varied User-Created Morphologies},
 booktitle = {Proceedings of ACM SIGGRAPH '08},
 note={\url{http://chrishecker.com/Real-time_Motion_Retargeting_to_Highly_Varied_User-Created_Morphologies}},
 year = {2008} }

http://chrishecker.com/Real-time_Motion ... rphologies

Code: Select all

@InProceedings{Poirier:2009:GI,
  author =       "Martin Poirier and Eric Paquette",
  title =        "Rig Retargeting for 3D Animation",
  booktitle =    "Graphics Interface 2009 Conference Proceedings",
  pages =        "103-110",
  year =         "2009",
}

http://profs.etsmtl.ca/epaquette/Resear ... rier.2009/

[ThomasL]

I have been looking into some papers by Michael Gleicher, http://pages.cs.wisc.edu/~gleicher/Papers , especially those from the later 1990's. In general most academics in CS have homepages with pdf copies of their papers.

The mocap tool currently in svn makes a rather simplistic approach to retargeting, which works reasonably well for realistic characters:

1. For most bones, I just copy joint rotations.

2. The root (hip, pelvis, com, ...) bone defines the overall rotation and location of the character. This must be the same, even if the source and target rigs have differently oriented and sized roots.

3. End effectors (IK hands, IK legs) follow the global location of their source counterparts.

To get a good result, one should help the tool by importing the bvh file with the right scale. I think the optimal scale is such that the vertical distance between hands and feet are the same, at least if you intend to use IK controls.

[ppClarity]

I've seen his '98 paper before, or at least the pictures I started to reread it last night and for my purposes (game), seems to be what I'm looking for. I'll need to work my way through the 2nd half where he discusses implementation.

I'd love to get my hands on this implementation (Wolfire's Overgrowth) for use in my own code:
http://blog.wolfire.com/2010/07/Overgro ... Animations