How to use classifier model that trained by ourself not tao to deepstream

how to use classifier model that trained by ourself not tao to deepstream.
how to write sgie_config.txt and infer code.

You can refer to our deepstream_tao_apps . In this demo we showed the useage of a couple of classifier model.

if i used other classifier model ,how i get .so file that write on custom-lib-path.

When you use your own custom model, you need to write its custom post-processer.
When you run make, a .so file is generated.
You give that path to custom-lib-path.

In deepstream_tao_apps git repo, you can refer the post-processer code for reference and modify it for your own model.

Once you run make and .so file is generated, just add it to your config file.
You need to add two parameters in the config file,
parse-classifier-func-name & custom-lib-path

Refer this.

1 Like

thank you so much. i will study this.

The above example is for detection model but I am sure you can find some examples for classifier models too.

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