Hi there,

I am developing an AR project and I need to do some head tracking.

 

The aim is to replace people’s heads with a 3d mask model, but I would like to know the orientation of the head (and this is the trouble) so that the object don’t look like an image, and to achieve progress in the application,like for example, turning left and right, up and down (yes/no), using the head.

I’ve thought some solutions including the use of kinect, or openCv.

But kinect head joint orientation is the same as the body, although this was the most consistent solution.  

With openCv I can detect the position of the eyes and mouth, to get the head orientation, or have several haarcascades files for different head positions, like front, profile ¾.. But this aren’t consistent at all..

If someone can give others ideas or suggestions I will appreciate ;)

Thanks people!