Creating new GXF component for secondary inference

• Hardware Platform (GPU)
• DeepStream Version (6.4)
• NVIDIA GPU Driver Version (550.54.14)
• Issue Type( questions)

I want to create a custom Deepstream GXF component for licence plate recognition, just like
NvDsSecondaryCarColorClassifierModel.
Can someone nudge me in the right direction?
I know that we can build a Component by overriding Codelet. But I am guessing that is too low level for my usecase.
Please correct me if I am wrong.

Thank You

The NvDsInferenceExt — DeepStream documentation 6.4 documentation has been provided to configure different models if there is no special customization post-processing is needed.

If you are talking about the License Plate Recognition | NVIDIA NGC model in NGC, please refer to the NVIDIA-AI-IOT/deepstream_lpr_app: Sample app code for LPR deployment on DeepStream (github.com) for the correct DeepStream pipeline.

The NvDsInferenceExt has provided UI to configure the new models.

Please set correct configuration to “config-file-path”. And please make sure the path in the configuration file should be better the absolute path.

1 Like

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