NVTRACKER with bounding box initiation

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

• Hardware Platform (Jetson Xavier NX)
• DeepStream Version 5.0
• JetPack Version (valid for Jetson only) 4.4
• TensorRT Version 7.1.3
• NVIDIA GPU Driver Version (valid for GPU only)

Is it possible to us NVTRACKER in the following way?
I would like to draw an ROI around an object to track in the image.
I would then like to continue tracking that object through subsequent frames (not necessarily using output from any object detection).

It this possible?
I tried using the KLT trackier in the VPI SDK but that is only available in C and I have to use Python.

Thank you!

3

@gbenel

May be you have to implement a simple GST plugin to set NvDsObjectMeta structure similar to the way inference (NvDsInfer) component fills it and then tracker will operate on this information to track the object.

More information about NvDsObjectMeta : NVIDIA DeepStream SDK API Reference: _NvDsObjectMeta Struct Reference

4

Thanks for the reply @ersheng .
Could you point to any examples of creating such a plugin?

I really wish that NVIDIA could work with openCV and python!

5

@gbenel

This is document about how to implement a new DS plugin.
https://docs.nvidia.com/metropolis/deepstream/dev-guide/index.html#page/DeepStream_Development_Guide/deepstream_custom_plugin.html#

And this is a sample plugin that you can use as a reference:
<location_of_deepstream>/deepstream-5.0/sources/gst-plugins/gst-dsexample