How to write the custom parser for the standard detect net output model?

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• GTX1070
• DeepStream 5.1
• TensorRT 7.2.3
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)

I have a detect net model with the output-blob-names=output_cov/Sigmoid;output_bbox/BiasAdd on its config file.

I checked https://forums.developer.nvidia.com/t/detectnet-custom-parser/107303

However it seems this is a bit old that it is for deepstream 3.2 and it doesn’t work with deepstream 5.1 at all.

I just need the standard default parser for the detect net and I need to modify it by just adding a little bit functions. How can I get the default parser’s code?

3

You can check nvdsinfer_context_impl_output_parsing.cpp → parseBoundingBox

4

@bcao
Should I just copy it directly? It seems needs some data structure from the header file nvdsinfer_context_impl.h

I checked the code they used the class DetectPostprocessor.
Should the parser only be one function directly? That is all other parser I used for pgie and siges.

5

the default built-in parser for detection model is nvdsinfer_context_impl_output_parsing.cpp → parseBoundingBox .
All these source code is open sourced, please check further