Delay in deepstream 6.3 pipeline

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) X86 machine with GPU
• DeepStream Version 6.3
• JetPack Version (valid for Jetson only)
• TensorRT Version 8.5.3.1
• NVIDIA GPU Driver Version (valid for GPU only) 535.171.04
• Issue Type( questions, new requirements, bugs) questions
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)

• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)

I have a deepstream6.3 python pipeline used for realtime object detection for multiple rtsp inputs. In the pipeline I am using nvstreamdemux and rtspclientsink to get AI stream output. I have been noticing a 2 seconds delay in the pipeline. Through various trial & error methods I have found that the nvstreamdemux is causing the 2 seconds delay and if I remove that element the detections are almost instantaneous. For my requirement, I need to use the nvstreamdemux element and I have to work with multiple (3) rtsp inputs(1280x720p with 25fps). Are there some nvstreamdemux properties that I am missing causing this delay?

3

Could you attach your whole pipeline and the various trial & error methods you have tried?

4

This is the pipeline

    streammux.link(nvvideoconvert)
    nvvideoconvert.link(capsfilter)
    capsfilter.link(pgie)
    pgie.link(tracker)
    tracker.link(nvanalytics)
    nvanalytics.link(fakesink)

I have removed the streamdemux element from the pipeline. It was used to get AI Stream of each rtsp stream to a front-end application. Because of the delay, I have removed the streamdemux and sink (rtspclientsink) elements and replaced it with fakesink.

5

OK. The delay may also be caused by the rtspclientsink. How do you get the delay after you use fakesink?

6

I’m not using the rtspclientsink all the time. By default, the fakesink is used and I’m switching to rtspclientsink only for a short time when the AI stream is requested from the frontend application. When the time limit exceeds, it will switch back to fakesink.

But the nvstreamdemux element is there all the time unlike the rtspclientsink. So I’m guessing its causing the delay issue. Maybe its buffering the frames for each stream? Furthermore, for single stream there is no delay. When we have multiple streams there is delay with nvstreamdemux

Can we use the add/remove nvstreamdemux dynamically like I have been using the fakesink & rtspclientsink? Or is there some property that can be used to avoid this delay?

7

You can try to set drop-on-latency to the rtsp source and configure the right value to the batch-size and live-sourceof nvstreammux.
Could you provide us a gst-launch-1.0 command to reproduce that issue? We can analyze that on our side.

8

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

9

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