none
Does Kinect v1 and background removal use the GPU? RRS feed

  • Question

  • Does anybody know what parts of the Kinect data processing is using the GPU?
    Does background removal use the GPU?
    In other words would a better GPU result in a better background removal performance?

    regards,

    Roland Smeenk

    Thursday, January 16, 2014 1:42 PM

Answers

  • The only features that rely on the GPU is KinectFusion. All other buffers are stored in memory and provided as RAW buffers. Depending on what the end result you are looking for, if you copy and move data to the GPU and back that can be more costly then just parallel computations on the CPU.

    Some of the latency you see has to do with getting the player/skeleton detection that has to be processed to get the PlayerMask values for the depth.


    Carmine Sirignano - MSFT

    Monday, January 20, 2014 9:38 PM

All replies

  • The only features that rely on the GPU is KinectFusion. All other buffers are stored in memory and provided as RAW buffers. Depending on what the end result you are looking for, if you copy and move data to the GPU and back that can be more costly then just parallel computations on the CPU.

    Some of the latency you see has to do with getting the player/skeleton detection that has to be processed to get the PlayerMask values for the depth.


    Carmine Sirignano - MSFT

    Monday, January 20, 2014 9:38 PM
  • Thanks, that was what I already expected. A better CPU helps more than a dedicated GPU.

    Roland Smeenk

    Monday, January 20, 2014 10:56 PM