How to output peoplenet inference results to a text file


I can run the peoplenet sample software as below.

$ cd /opt/nvidia/deepstream/deepstream-6.1/samples/configs/tao_pretrained_models/
$ sudo deepstream-app -c deepstream_app_source1_peoplenet.txt

As a next step, I want to output the bunting box coordinates inferred by peoplenet to a text file.
Please tell me how to do it. I searched, but I didn’t understand at all.

  • Hardware Platform (Jetson / GPU)
  • DeepStream Version
  • JetPack Version (valid for Jetson only)
    5.0.2 (L4T 35.1.0)
  • TensorRT Version

nvinfer plugin is opensource, you can output bboxes to a file in attach_metadata_detector of opt\nvidia\deepstream\deepstream\sources\gst-plugins\gst-nvinfer\gstnvinfer_meta_utils.cpp, please rebuild it and replace /opt/nvidia/deepstream/deepstream/lib/gst-plugins/


Thank you for your advice.

Does it mean that the attach_metadata_detector function can obtain the coordinates surrounding the person inferred by peoplenet?

Since the coordinates of the bounding box can be obtained with the attach_metadata_detector function, do you mean that you modify the source code so that you can save it in text yourself?

yes, please refer to
rect_params.left = obj.left; =;
rect_params.width = obj.width;
rect_params.height = obj.height;

yes, nvinfer plugin is opensource, please refer to the last comment.


I’m happy to get advice so quickly.

step1. I modify gstnvinfer_meta_utils.cpp and make.
step2. in lib directory is updated.

step3. I run peoplenet.

$ sudo deepstream-app -c deepstream_app_source1_peoplenet.txt

There is something I do not understand. How are you configuring it to use

please refer to my first reply, replace with the new one.

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