Deepstream Multistream slower than single stream

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU)
• DeepStream Version
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type( questions, new requirements, bugs)
• 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)

Hi, I am running the LPR deepstream SDK and i found out that when i run only 1 one-minute video, the application able to finish processing the video within 1 minute, but when i run 2 one-minute videos together (multistream), the application uses approximately 1 minute and 30 seconds. Not sure at which part of the pipeline is causing the delay.

Both video having 25fps, batch size for nvstreammux and pgie is set as the same number of sources, muxer batch timeout set as 40000.

Any idea on how to solve this because if i run on a live stream, the delay may be significant.

3

Please provide complete information as applicable to your setup. Thanks
Hardware Platform (Jetson / GPU)
DeepStream Version
JetPack Version (valid for Jetson only)
TensorRT Version
NVIDIA GPU Driver Version (valid for GPU only)
Issue Type( questions, new requirements, bugs)
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)

You can try to run the command below to test the performance of your 2 engine files on your board.

trtexec --loadEngine=xxx.engine <--int8 or --fp16> --batch=<1 or 2>
4

Hi, thanks for your suggestion. After working around, i found out that the main issue of my multistream having a delay is due to the resolution. However, the original resolution was 1920x1080, and if i decrease it down to 1280x720, it performs faster but the downside of it is having a lot of inaccurate results. If i would like to maintain my resolution (1920x1080), and I have tried increasing my interval to 16 and buffer-pool-size to 16, what are the other ways so that my application could load a smooth video

5

Besides, is it recommend to use drop-frame-interval from the decoder?

6

Could you attach the infomation below first?
• Hardware Platform (Jetson / GPU)
• DeepStream Version
It may be limited by the performance of your machine. You can check the performace of the model first with the command I attached.

Yes. You can use that if your machine is is heavily loaded.

7

I have try using the trtexec but results in segmentation fault.

[06/19/2024-16:52:00] [I] TensorRT version: 8.6.1
[06/19/2024-16:52:00] [I] Loading standard plugins
[06/19/2024-16:52:00] [I] Engine loaded in 0.0228568 sec.
Segmentation fault (core dumped)

Please advise

8

Since you can successfully run our LPR demo, this may be caused by your wrong command line.

Could you attach the whole command you are using, including the command to run our LPR demo?

9

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

10

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