Custom input of SGIE

[lqdisme]

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Jetson
• DeepStream Version 6.3
• JetPack Version (valid for Jetson only) 5.1.2
• Issue Type( questions, new requirements, bugs) questions

In my Python application, the current pipeline is structured as streammux → pgie (person detection) → sgie (object detection), where sgie operates on the bounding boxes provided by pgie. However, I would like sgie to receive expanded bounding boxes (i.e., enlarging pgie’s output to include more context). What is the solution for this, and could you provide an example?

[fanzh]

1. why do you need to expand the bboxes? how much do you need to expand the bboxes?
2. you can loop all object_meta generated by pgie, then modify the bbox rectangle coordinates. please refer to deepstream-test1 for how to loop object meta and deepstream-infer-tensor-meta-test for how to modify the bbox rectangle coordinates.

[lqdisme]

Thanks in advance. My solution is to write a custom gst-nvinfer with some custom configurations that can expand the primary GIE output bounding box before feeding it to the secondary GIE.

[fanzh]

Sorry for the late reply, Is this still an DeepStream issue to support? Thanks!

[yingliu]

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks