Queries about deepstream_parallel_infer_app

[lmw0320]

I just modified the code of deepstream_parallel_infer_app.cpp in order to save original images for each stream, which I can mark some info on it for callback。Also I modified the msgconv code to relate with the frame. But I found the duplicated message from the Kafka. And I found little time different with the message timestamp and frame timestamp.

1. I want to know why the messages are duplicated?
2. if possible, how to transfer the correct frame timestamp to message, which I can link the message with the correct frame image.

Here are the update save img code and msgconv code as below:

update msgconv code :

I also herein attach with the result of message and image savings :

[lmw0320]

From the snapshoot of image savings, we also can find that the fps are only 15 (but it should be 25fps in camera setting).
And the frame 0 was missing in source_id 0, but can be found in source_id 1.

[junshengy]

@lmw0320

I want to know why the messages are duplicated?

This is actually a feature of Kafka. It is usually caused by producer retries.

You can get more answers and solutions here or here

@lmw0320

if possible, how to transfer the correct frame timestamp to message, which I can link the message with the correct frame image.

I think you need to increase the time precision to milliseconds or microseconds to distinguish each batch

In addition, did you set the attach-sys-ts property of nvstreammux to false?
If set to TRUE, system timestamp will be attached as ntp timestamp. If set to FALSE, ntp timestamp from rtspsrc, if available, will be attached.

[junshengy]

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)
• The pipeline being used

[lmw0320]

1. We have both testing the code on Jetson and GPU. and the Deepstream version is 6.2 with Tensorrt version 8.5.2.2.

2. for the frame timestamp, I just want to confirm the kafka message to be linked with correct frame number which I may use the correct image for usage.

3. I also want to know why the saving images only 15 for 25fps camera setting?

4. And why the frame number of 0 is missing for source_id 0 ?

[junshengy]

@lmw0320

I also want to know why the saving images only 15 for 25fps camera setting?

Please make sure the camera outputs 25 images per seconds.
I guess you are using v4l2src element, For v4l2src, need to copy data from cpu to gpu in deepstream_source_bin.c. This will impact performance

In addition, using OpenCV to save images requires copying data from the GPU to the CPU, which will also affect the performance of the pipeline.
Better practice, Refer to the usage of nvds_obj_enc_process in deepstream-image-meta-test.

@lmw0320

And why the frame number of 0 is missing for source_id 0 ?

Why is the parameter *frame_num? Shouldn’t it be frame_num[source_id]?

[lmw0320]

Thanks. I will check the code first.
for the saveing image function, It seems I used wrong parameters… I would modify and recheck.

[junshengy]

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