[Bf-committers] GSoC proposal

[Sergey Kurdakov]

Hi Shuvro,

I believe, that there are could actually many directions in implementing
what I mentioned in previous message.

1)st auto calibrate kinect(s) see available software
http://graphics.cs.msu.ru/en/science/research/calibration/cpp
2)a reconstruct scene from kinect using  http://vimeo.com/21096739 ( diff
here http://www.pasteall.org/20117/diff )
using either Kreilos approach or maybe
http://www.ros.org/wiki/openni/Contests/ROS%203D/RGBD-6D-SLAM approach
<http://graphics.cs.msu.ru/en/science/research/calibration/cpp>2) b fit
generic figures to captured data ( generic figures could be taken from
MakeHuman, I think )
3) estimate motion. see additionally link
http://www.ros.org/wiki/openni/Contests/ROS%203D/Skeleton%20Tracker%20Teleoperation%20Package%20for%20Mobile%20Robot

<http://www.ros.org/wiki/openni/Contests/ROS%203D/Skeleton%20Tracker%20Teleoperation%20Package%20for%20Mobile%20Robot>so
pretty much to do.

but unlike your proposal - it does not rely on patented pending approach
implemented in Nite based code (Extraction of skeletons from 3D maps United
States Patent Application 20110052006 ) and potentially provides much more
reliable
motion data capture.

as for Blender and Nite you may see here
http://www.youtube.com/watch?v=UxIcwuo5Rts it is already tested with
http://www.brekel.com/  tool - which is basically a wrapper around Nite.

I just add some info to think on if the way you describe is really worth
pursuite

Regards
Sergey