none
dropping color and depth frames RRS feed

  • Question

  • Here are some results I'm getting from my camera and  code. I get the same results when I look at the frame numbers in the Skeletal Viewer sample. Is this typical? Or can it do better. It looks like one frame in 20 is dropped or so is dropped, both depth and color.

    color depth   col time  dept time   dif

       838    838   27.890   27.900   -0.010
       839    839   27.920   27.930   -0.010
       840    840   27.950   27.960   -0.010
       841    841   27.980   27.990   -0.010
       842    842   28.020   28.030   -0.010
       843    843   28.050   28.060   -0.010
                 844                 28.090
       844    845   28.110   28.120   -0.010
       845    846   28.140   28.150   -0.010
       846    847   28.180   28.190   -0.010
       847             28.210
       848    848   28.240   28.250   -0.010
       849    849   28.270   28.280   -0.010
       850    850   28.300   28.310   -0.010
       851    851   28.340   28.350   -0.010
       852    852   28.370   28.380   -0.010
       853    853   28.400   28.410   -0.010
       854    854   28.430   28.440   -0.010
       855    855   28.460   28.470   -0.010
       856    856   28.500   28.510   -0.010
       857    857   28.530   28.540   -0.010
       858    858   28.560   28.570   -0.010
       859    859   28.590   28.600   -0.010
       860    860   28.620   28.630   -0.010
       861    861   28.660   28.670   -0.010
       862    862   28.690   28.700   -0.010
       863    863   28.720   28.730   -0.010
       864    864   28.750   28.760   -0.010
       865    865   28.780   28.790   -0.010
       866    866   28.820   28.830   -0.010
       867    867   28.850   28.860   -0.010
                 868                 28.890
       868    869   28.910   28.920   -0.010
       869    870   28.940   28.950   -0.010
       870            28.980
       871    871   29.010   29.020   -0.010
       872    872   29.040   29.050   -0.010
       873    873   29.070   29.080   -0.010
       874    874   29.100   29.110   -0.010
       875    875   29.140   29.150   -0.010

    This is a typical run. Sometimes it goes for longer without dropping frames. Can the Kinect do better?

    Michael

    Monday, May 7, 2012 3:21 AM

Answers

  • Not quite sure what your question is. What are you looking to do? When capturing 30fps, you can experience a dropped frame every second. This is going to be subjective based on the system. What type of system load are you experiencing at the time you see the frame drop. Do you have other processes running when doing video capture. Is this debug or release mode?

    If you are using an event/callback type event system (Skeleton viewer) there is no guarntee that when you query the data, that the depth and color data will be in sync. If you need to make sure color and depth data are for the same frame, in-sync, use the pull model.

    • Marked as answer by stereosphere Tuesday, May 8, 2012 9:56 PM
    Monday, May 7, 2012 11:49 PM

All replies

  • My Kinects drop a depth frame about once every thousand frames. Is this normal? I get this behavior with my code and with a slightly modified version of the skeletal viewer.

    Thanks.

    Wednesday, April 25, 2012 5:38 PM
  • Here is what actually happens (I think I've got it right) :

    color depth   col time  dept time   dif

       838    838   27.890   27.900   -0.010
       839    839   27.920   27.930   -0.010
       840    840   27.950   27.960   -0.010
       841    841   27.980   27.990   -0.010
       842    842   28.020   28.030   -0.010
       843    843   28.050   28.060   -0.010
                 844                 28.090
       844    845   28.110   28.120   -0.010
       845    846   28.140   28.150   -0.010
       846    847   28.180   28.190   -0.010
       847             28.210
       848    848   28.240   28.250   -0.010
       849    849   28.270   28.280   -0.010
       850    850   28.300   28.310   -0.010
       851    851   28.340   28.350   -0.010
       852    852   28.370   28.380   -0.010
       853    853   28.400   28.410   -0.010
       854    854   28.430   28.440   -0.010
       855    855   28.460   28.470   -0.010
       856    856   28.500   28.510   -0.010
       857    857   28.530   28.540   -0.010
       858    858   28.560   28.570   -0.010
       859    859   28.590   28.600   -0.010
       860    860   28.620   28.630   -0.010
       861    861   28.660   28.670   -0.010
       862    862   28.690   28.700   -0.010
       863    863   28.720   28.730   -0.010
       864    864   28.750   28.760   -0.010
       865    865   28.780   28.790   -0.010
       866    866   28.820   28.830   -0.010
       867    867   28.850   28.860   -0.010
                 868                 28.890
       868    869   28.910   28.920   -0.010
       869    870   28.940   28.950   -0.010
       870            28.980
       871    871   29.010   29.020   -0.010
       872    872   29.040   29.050   -0.010
       873    873   29.070   29.080   -0.010
       874    874   29.100   29.110   -0.010
       875    875   29.140   29.150   -0.010

    This is a typical run. Sometimes it goes for longer without dropping frames. Can I trust these numbers? Can the Kinect do better?

    Michael

       

    Wednesday, April 25, 2012 11:25 PM
  • Not quite sure what your question is. What are you looking to do? When capturing 30fps, you can experience a dropped frame every second. This is going to be subjective based on the system. What type of system load are you experiencing at the time you see the frame drop. Do you have other processes running when doing video capture. Is this debug or release mode?

    If you are using an event/callback type event system (Skeleton viewer) there is no guarntee that when you query the data, that the depth and color data will be in sync. If you need to make sure color and depth data are for the same frame, in-sync, use the pull model.

    • Marked as answer by stereosphere Tuesday, May 8, 2012 9:56 PM
    Monday, May 7, 2012 11:49 PM
  • Thanks, that answers my question. I was asking if the behavior I am seeing is normal, or if the Kinect can do better. 

    Michael

    Tuesday, May 8, 2012 10:22 PM