How to get a unique key of each frame?

[colocasia98]

• Hardware Platform: Jetson
• DeepStream Version: v6.1(docker container)
• JetPack Version: 5.0.1 DP [L4T 34.1.1]
• TensorRT Version: 8.4.0-1+cuda11.4
• Issue Type: questions

Hi all,

I tried to run multiple deepstream pipelines which have same rtsp input, to achieve redundancy.
I found that the “frame_num” in frame metadata starts from 0 in each pipeline. When I ran 2 pipelines at same time, one frame always had two different “frame_num” due to different start times of the two pipelines.
So, I’m wondering if there’s any key value (or any other thing) of a frame so that I could specify each frame and de-duplicate the results when I run multiple pipelines?
Thanks.

[yuweiw]

What do you mean by redundancy?
You can try to use the timestamp of your rtsp source to specify each frame.

[colocasia98]

Thank you for your reply.

The redundancy here means that I would like to run the pipeline in two or more physical machines.

I tried to fetch the timestamp by using frame_meta.ntp_timestamp
but it was still not same in different machines.

I would be really appreciating if you could give more information about how to get the timestamp of the rtsp source via deepstream.

Thanks.

[yuweiw]

You can refer to the streammux Guide below and try to set the attach-sys-ts para to False.
https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_plugin_gst-nvstreammux.html