Opencv head traking with webcam

im working on a project where kinect wont do the job, i need something like this…

i would like to know what are the first steps that i need to know on how to use opencv with vvvv to get the head position, any experience on this ?

you can use DetectObject (FreeFrame DShow9) for that
(the link and the nodes helppatch will offer more info)

This is 3d head tracking…
like this:

I don’t know detectObject can do this…

I’ve got nearly good result using haarcascade trained xml for face, eye, but in 2d.

I’m also interested in this topic,
How to achieve real 3d face tracking in vvvv?


EDIT: this could be useful…

there are many ways to rome ;)

One would be to use the commercial faceapi, another would be to use the (only noncommercial) FaceTracker Library by Jason Saragih. Anyway all of these are c++ solutions - you had first to wrap them in c# or as a freeframe for use these with vvvv - and that’s definitely a non trivial task.

sebl i wasnt able to run detect object, well, it open, i can see my webcam
i already try with several haarcascade_frontalface_alt.xml but i dont get nothing but vvvv crashes when i load them

but as robe said, thats a 2d traking, i need a 3d traking…
i like the m9dfukc solution, i will try it on my mac latter, if i have the osc message i send that to vvvv and thats it, but i will be lovly to get the same under vvvv or on another app under windows, well procesing seems to be the answer, the combination vvvv + procesing connected trough osc seems a good workarround… i will post my progress ( if there is any :P ) here

you can also use Kyle McDonalds ofxFaceTracker which is a openframeworks wrapper around the FaceTracker lib mentioned above. In ofxFaceTracker a osx messaging system is already implemented sensing head pose and facial feature data. … precompiled libs should available here