When we using detection, how extract the position of detections boxs?

When the detection running, can we extract the detection position, to use them with an other programe than drawing the boxs on the picture?

I use a Jetson NX and the last version of JetPack (4.5.1)

We suggest try DeepStream SDK. In DeepStream SDK, we have nvdsosd plugin to draw the rectangles. You may check the document:
Gst-nvdsosd — DeepStream 5.1 Release documentation
And see if you can use it in your usecase.

If you would like to have your own implementation, you can get metadata and implement a custom plugin.

I’m sorry to reply so slowly.
I guess I have wrongly formulated my question: I am not want to draw the boxes, but extract the box’s datas to use them at something else than drawing boxes.
For the moment I make detection box by following this tutorial (I should have told it in the original question) : GitHub - dusty-nv/jetson-inference: Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson. .