bit.ly/​bWBfSh

is the link @silicaroach mentions

it a thinning algo. attempting port to processing now.

asked theo about his extracted contour analysis approach. i think its quite specific to the one-arm-puppet thing...

Oh, awesome, that tutorial looks helpful.  I would like to somehow integrate skeleton tracking with my library, but I'm using openkinect instead of openni.  OSC is obviously a great way to go, but what a pain to install OPENNI, etc.!  This is definitely next on my list if anyone has any suggestions or helpful advice, please be in touch!

Amazing, fantastic work!   If there's anything I can do to help or if you have any thoughts on ways that my "openkinect" library and the "openni" library could work together, let me know.  Really excited to play with this more. . 

Suggestion: even if it is just a Google search away, put a prominent link to OpenNI (personally, I don't know this library) on the pages of your project.

Thanks for this awesome library. Yours is actually the fourth Kinect library have used (after Shiffman's, nrocy's and OSCeleton). I am switching to it as it seems to get rid of the segmentation faults that I was getting with OSCeleton. It is handy to have access to both the skeleton and the raw depth image.

I am just writing about a couple of issues that I have come across using version 0.11

The first is that if I use OPENGL with depthMap generation switched on with enableDepth() and accessed with context.depthImage(), I get a nullPointer exception (P3D mode doesn't work either). Is there a way to use OPENGL with the depthMap? The application I am writing is a motion capture/replay system that allows children to record the moves of their clapping games and play against their recording and I need OPENGL so that I can draw a pair of oversized hands on the skeleton and rotate them appropriately (I am using a Kinect to track the player and Wiimotes to track their hand orientation).

I have also noticed that the depth map appears to have vertical lines running through it - what it odd about these is that they get closer together the longer the sketch runs. I am working on a MacBook Pro.

BTW - is there a function like getJointPositionSkeleton() which just returns a PVector, rather than requiring me to pass in a PVector that it sets? This would be useful as I am just passing the joint coordinates straight into a setXYZ() function.

Thanks for the nice comments ! But the OpenNI library makes the hole magic, i just write a wrapper around it.

@shiffman

Help is appreciated, on the mac, the hole installation of  openNI(before you can use SimpleOpenNI) is quite time consuming. It would be nice to have a pkg for osx, but i'm not that experienced with osx.

About the two libs, i actually tried to make the functions look similar.

Updated the page, the links to openNI are now on the start page.

Do you have an example source ? I quick tested it with SampleOpenNI_DepthImage, i set the rendering to opengl, but it worked(v0.12).

About the stripes in the image, i also noticed this, but never that strong, have to check this more in detail. But the depthMapImage is more for the visual check, otherwise i would use the depthMap raw data(gives the depth in millimeter) or even the depthMapRealWorld(the real 3d data).

About getJointPositionSkeleton(), this function gives a float as return, this value should reflect the confidence of this joint, sounds strange. But maybe i should make this function for simplicity.

By the way, i just release a new version:

Do you mind if we would shift the discussion about an installer to the simpleOpenNI forum ?