none
C++ Face Tracking Error FT_ERROR_FACE_DETECTOR_FAILED RRS feed

  • Question

  • Hi,

    I'm adding the Face Tracking support to an existing tool and I'm not able to get the tracker to start successfully.

    When calling StartTracking(...) the return HRESULT is S_OK but the GetStatus() always return the FT_ERROR_FACE_DETECTOR_FAILED error.

    My tool is supporting Multi Kinect usage so all the Inits for Sensor, Color, Depth and Skeleton are associated with the active KinectSensor Instance(s). The FTCreateFaceTracker() doesn't seems to hold any information of which kinect is active.

    Although I support multi kinect in the tool, I am currently testing with only one...

    It look and found nowhere that the FaceTracker is not supported in a Multi Kinect environment.

    Thanks

    Monday, May 28, 2012 8:12 PM

Answers

  • The face tracker instance can be fed with RGB and depth data from any camera. It is not attached to a particular Kinect camera (but you need at least one Kinect camera to be attached to a PC in order for the Face Tracking API to function).

    FT_ERROR_FACE_DETECTOR_FAILED error means that the internal face detector failed to find any face on an RGB frame that you are passing to StartTracking() method. I would recommend writing out the RGB frame as an image and checking if it is correct (i.e. read from the right camera and not blank). Is it possible that you did not fill the IFTImage interface with the right RGB data from the working camera? If everything is correct code wise, then the issue could be with the amount of light or the camera orientation - the face detector will not be able to detect faces if an RGB frame is too dark or if the camera is too low or too high or too much to the side relative to a face. 

    Check this article - http://nsmoly.wordpress.com/2012/05/21/face-tracking-sdk-in-kinect-for-windows-1-5/  for coding tips and other recommendations. The code sample shows the proper API usage and the tips at the bottom describe possible camera related issues.

    Wednesday, May 30, 2012 6:20 AM

All replies

  • The face tracker instance can be fed with RGB and depth data from any camera. It is not attached to a particular Kinect camera (but you need at least one Kinect camera to be attached to a PC in order for the Face Tracking API to function).

    FT_ERROR_FACE_DETECTOR_FAILED error means that the internal face detector failed to find any face on an RGB frame that you are passing to StartTracking() method. I would recommend writing out the RGB frame as an image and checking if it is correct (i.e. read from the right camera and not blank). Is it possible that you did not fill the IFTImage interface with the right RGB data from the working camera? If everything is correct code wise, then the issue could be with the amount of light or the camera orientation - the face detector will not be able to detect faces if an RGB frame is too dark or if the camera is too low or too high or too much to the side relative to a face. 

    Check this article - http://nsmoly.wordpress.com/2012/05/21/face-tracking-sdk-in-kinect-for-windows-1-5/  for coding tips and other recommendations. The code sample shows the proper API usage and the tips at the bottom describe possible camera related issues.

    Wednesday, May 30, 2012 6:20 AM
  • In addition to what Nikolai said, I would recommend playing with the SingleFace sample to get a feel for how the API works.

    Also, multiple sensor scenarios should work fine.  The FT components don't care which Kinect sensor the frames are coming from.  The application passes the frames from an arbitrary Kinect sensor to the FT components.

    Wednesday, May 30, 2012 1:52 PM
  • I haven't forgotten the thread!!! :)

    Thanks for your input to both.

     
    I think you are right about the image just not being correct. I've changed the buffer to allocate based on the samples but I had to change the format to G8B8R8X8 which allow me to properly see the data now but the GetStatus() now give me E_UNEXPECTED so I'll have to give it a look again.

     
    Unfortunately I have to post-pone the work on that side but I will get into it again and will give feedback at that time.

     
    The built samples work fine so I am not questionning the lighting, camera setup, distance, device...

     
    Thanks

    Monday, June 11, 2012 2:10 AM