NvStreamMux (New) adaptive-batching and NvTracker

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Jetson
• DeepStream Version 6.3
• JetPack Version (valid for Jetson only) JetPack 5.1.2 GA
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type( questions, new requirements, bugs)
Seems to be a bug in the tracker reference implementation???
• 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)

I’m running into the exact same issue as posted here

If I set adaptive-batching to true and start the Pipeline with a single source. When you add a new Source the Tracker reports

'Frame number in input batch exceeds maximum batch size'

Which results in additional errors

0:00:06.723566286 27459     0x30339700 WARN                 nvinfer gstnvinfer.cpp:2397:gst_nvinfer_output_loop:<infer-primary-gie-nvinfer> error: Internal data stream error.
0:00:06.723613711 27459     0x30339700 WARN                 nvinfer gstnvinfer.cpp:2397:gst_nvinfer_output_loop:<infer-primary-gie-nvinfer> error: streaming stopped, reason error (-5)

If I remove the Tracker plugin from the Pipeline everything seems to work ok. The only workaround is to set the Streammux gst-property to the max-sources-to add.

I would think you could easily reproduce this with your https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/tree/master/runtime_source_add_delete example.

Thanks,
Robert.

nvtracker will allocate the memory based on the batch size. Please set the Streammux gst-property to the max-sources-to add.

Hi, @kesong
I have one question.
What is the exact property name of Streammux that should be set as the max sources count?

Hi, @kesong
I have one question.
What is the exact property name of Streammux that should be set as the max sources count?

@sieman216485 The GST property is called batch-size … see here

… and FYI. based on my experience, once the solution box is checked on a topic as above, best to open a new topic if you want NVIDIA to respond.

@prominence_ai
Thanks for your reply.

I already tried setting batch-size but still face the same issue.
Anyway, I will open a new topic to get NVIDIA’s response.

Thanks for your tip.

@sieman216485 make that you’re not setting the batch-size in the Streammux config file as well… as this will override the GST property setting. Or, if you do set it in the config, set it to max-sources-to-add here.

From the docs…

“This value will override plugin property and DS config file key “batch-size” for nvstreammux
If batch-size not specified in the config-file, plugin property batch-size shall override the default.”

Hi, @prominence_ai Thanks again for your reply.

After testing different cases, I confirmed that it only works when I set gst-property, regardless of whether I set the config-file property.

However, according to the documentation, the config-file property value will override the gst-property.

But my testing result shows that it is not.

What do you think about this?

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