How does the Gst-nvtracker's state estimator get and use time to calculate derivatives like dx, dy, dw, dh

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Hello.

I am hoping you can help me understand how the low level tracking libraries calculate the derivatives used in the state estimator (dx, dy, dw, dh). I understand the tracking libraries are closed source, but would you be able to tell me how they get the time component needed to perform the derivation. Do they use GstClock? Is it something more simple like getting the FPS of the source?

The motivation for knowing this is:

  • If I was to implement a custom tracker how could I get time to calculate my own derivatives?
  • In my testing pipeline can I run videos sped up without impacting tracker performance?

Thank you

3

Here is the guide for nvtracker state estimator: Gst-nvtracker — DeepStream documentation 6.4 documentation
Different low level library have different state extimator implementation. You can implement your own state extimator. As there are use case when timestamp is unavailable, dx dy may the diff of x and y between frames.

4

Hi Kesong

Thanks for your reply, that documentation was how I came to asking this question. So if in NvDCF the derivatives are just calculate as the differences between frames then speeding up my videos in my test pipeline should not change tracker performance. Provided FPS is also increased in proportion.

Does that sound correct? NvDCF does not use any form of real time from the host system?

5

Yes, you are right.

6

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.