Time consuming in tx2

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

Hi all.
I had encountered a time consuming problems of video object detection.First, we decode the h.264 format video file,then
convert YUV image to RGBA image. When running 1 thread object detection using gpu device,the average consuming time is about 90 ms per frame. but the consuming time of fluctuation is between 80 ms per frame and 400 ms per frame. When running 2 threads object detection sometime the consuming time will be less,and sometime it will be more.That is random.
So My questions are:

  1. What cause these issues？
  2. How to solve these problems？
  3. Does the graphical desktop of linux system will use GPU device？
    4）How to control to call GPU resources by user program application？
2

Please share detail about how you do h264 decoding → YUV to RGBA conversion → object detection. Do you implement it in MMAPIs or gstreamer?

3

Hi neon-1004,

Have you clarified and resolved the problem?
Any update?

Thanks