assurance-tunnel
assurance-tunnel
assurance-tunnel
assurance-tunnel

Ceiling tracking Research



  • Dear all,
    I'm working on a new installation where for the first time I'm going to track the participants movements with a kinect1 in the ceiling.
    During my test, I simulated white dots with After-effects to work in Isadora.
    When One dot appears in the screen, the tracking works well :
    https://vimeo.com/150163225

    But if I have 3 dots moving in the same or opposite directions it becomes a mess :
    https://vimeo.com/150163224
    When 2 dots come close to each other they get the same colour even if they move in opposite directions

    I was wondering if there was a way to fine tune the Eyes ++ for a better result?
    Or does anybody have hands on experience on this?

    Here's an image of my setup.

    Thank you for your help.
    86b278-ceilingtracking.jpg


  • Tech Staff

    @David

    This is one where you have to play around with "same max dist" and "same area change".

    If you set either of them to a high value they are basically disabled. Try to set "same area change" to 100 to disable it and start with a value of 5 for the "same max dist" and see what is happening.

    Best Michel



  • Hi @Michel,
    thank you so much for your feedback. I'm going to try this.
    Thank you!
    All the best.

    David


  • Tech Staff

    If you are using a kinect are you using the depth map image to get the tracking? Or are you using a third party software such as ni-mate?



  • Mmmm... if doing the dots out of Isadora, e.g.: with Processing or something more tweakable, then you can make the presence/individuality of dots depth-dependant as well as x / y dependant, and just send over the x/y co-ords to Isadora via OSC for rendering the effect...

    At the moment, working with an AE render, the issue is presenting itself regardless in 2D,but see whether it makes a difference by giving each of your particle renderers a different z translation (10 / 20 / 30 for instance)... if that corrects the issue then tracking z for each blob from the kinect and mapping it to z translate of the particle actors would probs also resolve the issue (??)


  • @Skulpture I'm using a Demo version of CocoKinect developped by Rob Sharein.
    It turnes the spectator into red filled form.

    @Marci, thank you! Pretty clever! I'm not good at processing yet. But I'll see how to implement these ideas.