Custom Parser for HRNet in Deepstream SDK 5.1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version 5.1
• JetPack Version (valid for Jetson only)
• TensorRT Version 7.0
• NVIDIA GPU Driver Version (valid for GPU only) 10.2
• 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 am trying to run inference on a live video stream using deepstream 5.1 sdk and want to use a HRNet which I have trained on my own dataset. HRNet is not one of the supported models in the deepstream SDK as per Using a Custom Model with DeepStream — DeepStream 5.1 Release documentation. Hence, I am stuck at this step as I am not proficient enough in C++ to write my own custom parser scripts.
The following are my questions:

  1. Are there any custom parser(.so) available out of the box which I can use for HRNet?
  2. Are there any tools available which let me generate the custom parsers for any model? along with the associated Deepstream configuration file?

do you mean a parser to parse the output of HRNet?

There is not such tool, and since different models have differents output formats, I think such tool may be not doable.
so, you need to write it by yourself according to the output format of HRNET.

Thanks!

@mchi yes by parser I mean the output parser. which I can use for the following parameters:

  • parse-bbox-func-name

  • custom-lib-path

Thanks for answering the second question. Are there any generic steps I can follow to create the custom parsers on my own?

you can refer to the post_processor under GitHub - NVIDIA-AI-IOT/deepstream_tlt_apps: Sample apps to demonstrate how to deploy models trained with TLT on DeepStream, which are the output parser for different kinds of models