At present, I need to be able to connect to the human key point model of deepstreeam5.0, but I am not familiar with it. Are there any recommended open models that have been connected?

• Hardware Platform (Jetson / GPU)
• DeepStream Version
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)


I am so sorry that we currently do not have similar samples as human face landmarks.
And further more, we do not have direct support of post-processing of landmark data types.

However, NvDsInfer is now opensource. You can find source code of NvDsInfer in .../sources/gst-plugins/gst-nvinfer (high level) and ../sources/libs/nvdsinfer (low level).

So, you can develop your own NvDsInfer by:

  1. customizing NvDsBatchMeta or any sub types inside NvDsBatchMeta like NvDsFrameMeta or NvDsObjectMeta
  2. creating a new landmark type (You can have a look at NvDsInferObjectDetectionInfo or NvDsInferAttribute as reference) to implement parsing for landmark tensor outputs.

Documents as references:

NvDsInfer Plugin introduction

NvDsInfer API (NvDsObjectMeta)