Scaling problem with GetProjectedShape function RRS feed

  • Question

  • Hello,

    Using Kinect SDK 1.8 (did not try with older version), I do Face tracking using RGB with 640x480 geometry and depth with the same geometry. When I have a face, I ask for the FaceRect and everything is ok. But when I ask for projected shape, there is a scaling problem. I check videoConfig param is ok, scale param is ok (1.0f) and other params where gatheres using IFTResult or IFTModel. I ran the sample code provided It works.

    Does someone experiment such kind of problem ? The sample code seems to use 320x240 for depth. Is it mandatory ?

    Thanks you in advance.


    Friday, December 20, 2013 10:29 AM


All replies

  • You can use 640x480 or 320x240, the main factor is speed of the recognition. If you system isn't fast enough you could get latency issues/slower recognition or failures. You can apply adjustments to the project results since these are just vector positions that you can manipulate with a matrix.

    Carmine Sirignano - MSFT

    Friday, January 3, 2014 1:01 AM
  • Hello,

    Thank you for your answer. My main concern is that the scaling of the projected face is not good at 640x480 (I did not experiment any speed problem). Indeed, the projected face do not match the face (it is out of the bounding box gathered from the Kinect).

    Best. Doms.

    Monday, January 6, 2014 9:32 AM
  • Using the Facetracking samples provided in the toolkit, what do you see? Can you provide a screen-shot?

    Carmine Sirignano - MSFT

    Tuesday, January 7, 2014 11:21 PM
  • Hello,

    As I said, the provided example works fine *but* it is using a 320*240 depth image, not 640*480. That's the point of my question. Until end of next week, I can not make a screen shot, sorry.


    Friday, January 10, 2014 3:58 PM
  • Did you change the sample to use 640x480 instead? The samples provide full source code so you can modify it to fit your needs.

    Carmine Sirignano - MSFT

    Saturday, January 11, 2014 1:16 AM