How optimize the custom logic in Deepstream python SDK

nareshmungpara · July 21, 2021, 11:24am

• Hardware Platform (Jetson / GPU) : Jetson Nano
• DeepStream Version: 5.1
• JetPack Version (valid for Jetson only) : 4.5.1
• TensorRT Version : 7.1.3.0

Hello there,

I am using deepstream_imagedata_multistream.py now I want to add some custom logic in this script.

I have already tried it but as soon as I add something fps drops drastically.
I am trying to change tiler_sink_pad_buffer_probe, changing anything in this function it increases the processing time and fps drops.

Is there any optimized way or method to do it?

Thanks,
Naresh Mungpara

AastaLLL · July 22, 2021, 2:53am

Hi,

Could you share the custom logic with us to check?
Since it is a multistream pipeline, please make sure your logic can support number_sources concurrently.

github.com

NVIDIA-AI-IOT/deepstream_python_apps/blob/master/apps/deepstream-imagedata-multistream/deepstream_imagedata-multistream.py#L293

    
      
          if not pipeline:
              sys.stderr.write(" Unable to create Pipeline \n")
          print("Creating streamux \n ")
          
          
# Create nvstreammux instance to form batches from one or more sources.
          streammux = Gst.ElementFactory.make("nvstreammux", "Stream-muxer")
          if not streammux:
              sys.stderr.write(" Unable to create NvStreamMux \n")
          
          
pipeline.add(streammux)
          for i in range(number_sources):
              os.mkdir(folder_name + "/stream_" + str(i))
              frame_count["stream_" + str(i)] = 0
              saved_count["stream_" + str(i)] = 0
              print("Creating source_bin ", i, " \n ")
              uri_name = args[i + 1]
              if uri_name.find("rtsp://") == 0:
                  is_live = True
              source_bin = create_source_bin(i, uri_name)
              if not source_bin:
                  sys.stderr.write("Unable to create source bin \n")

Thanks.

nareshmungpara · July 24, 2021, 9:58am

Hello,

While I was going through the code I have worked on num_source. Currently I am working object tracking and for that I have Implemented tracker part in the code for different num_source. Right now it’s working fine. But as it in tiler_sink_pad_buffer_probe function. And that’s why I need more optimized way to intergrate it.

Thank you,
Naresh Mungapara

AastaLLL · July 26, 2021, 7:32am

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

Hi,

Would you mind sharing the source of your customized tiler_sink_pad_buffer_probe function?
So we can check the performance issue directly?

Thanks.

system · October 9, 2021, 12:57am

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.