kinect tracking suggestion


    Mar 28 2013 | 1:51 pm
    hello max fellows,
    i want to start an interactive project with the kinect.
    the goal is to track 2 skeletons with my xbox kinect without doing the “psi” pose.
    did a research and found out, that are many possibilities to do this.
    mac versus win, openni versus mirosoft sdk, xbox versus windows kinect and so on.
    i have tested several apps, which send osc messages with the skeleton data.
    something as kinecta is what i am looking for:
    but it crashes often on my mac.
    i am looking for some hints from people with experience at this field.
    can someone suggest a simple and stable solution. mac or pc. (mac prefered)
    thanks in advance,
    pixelschub

    • Mar 28 2013 | 2:49 pm
      jit.openni for Max is great.
      SimpleOpenNI for Processing too.
    • Mar 28 2013 | 3:55 pm
    • Mar 29 2013 | 10:50 am
      thank you very much dtr and matmat for your hints !
      i think i will go the jit.openni route.
      dp.kinect is not an option, because it is not clear how much it will cost after the beta period.
      information from the dp.kinect website :
      "Current Beta version is v0.6.5 with an expiration date of 28 Mar 2013."
      pixelschub
    • Mar 29 2013 | 12:06 pm
      Hi everyone,
      I didn't think you could do skeleton tracking without the pose?
      I hope I am wrong on this as it would be great!
      can jit.openni do this?
    • Mar 29 2013 | 12:58 pm
      with jit.openni it is possible.
      i am doing it right now :-)
      does anybody has made a simple 3d model (jit.gl.sketch?? ) in jitter, which represents the skeleton coming out of jit.openni ?
      i have found: the Viewskeletons patch, but it is a bit different.
      http://www.cycling74.com/forums/topic.php?id=37199
      it is using ./osceleton as a input base.
      pixelschub
    • Mar 29 2013 | 1:10 pm
      What I do is simply put the skeleton coordinates in a 15x1 matrix and run that into a jit.gl.mesh in point drawmode. I don't need more visualization than that. I'd share it with you but it's embedded in a larger patch, no time to extract right now. Very easy to replicate though.
      Btw, I'd largely prefer jit.openni to output the joints as a matrix rather than the messages it does now.