[Bf-committers] GSoC proposal
sergey.forum at gmail.com
Sun Mar 27 10:14:19 CEST 2011
I have one thought to share on subject of motion capture
that while there are a lot of samples now when kinect allows to get motion
- one kinect is very hard to make to produce good animation data.
so, while there are quite a bit examples to use Kinect to make motion
capture - it exibits weakness here and there - captures mostly standing
persons, feet sleeps, etc.
There is another approach though.
To use two kinects.
You might look at Oliver Kreylos
http://idav.ucdavis.edu/~okreylos/ResDev/Kinect/ page and also
http://www.cs.ubc.ca/labs/imager/tr/2009/eltopo/eltopo.html software - then,
I expect, it is possible to reconstruct skeleton at each time frame.
To estimate mesh from Oliver Kreylos data you might use recently
implemented double contouring algorithm
see http://vimeo.com/21096739 ( diff here http://www.pasteall.org/20117/diff
this way - it would be possible to capture professional quality motion
capture data and not something with sleeping feet.
as a side note you may look at paper
while not directly related - it might be of some help in implementing above
On Sun, Mar 27, 2011 at 11:01 AM, shuvro sarker <shuvro05 at gmail.com> wrote:
> This is the link of my GSoC proposal. There are two links in the following
> page -
More information about the Bf-committers