Some details wanted about Secondary gie mode ?

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Hello,
I am working on cascade detector ,
like resnet10 cascaded with yolo detector,
but there were always a cuda memory problem,
I have read all the developper guide,
and found that the "Secondary gie mode " is designed for object detect (not full frame),
and the demo dtest1~dtest4 always use the secondary gie like :

…detect–> classifier1 → classifier2 —> classifier3–>…

and so on,

so I want to know if the secondary detector is supported by now,
and where is the demo resnet10 detected network could be got, I want to retrain it,

Any message is appreciated!

2

You can refer to https://github.com/vat-nvidia/deepstream-plugins to implement your own tensorRT gstreamer plugin.

Deepstream 4.0 will support Primary detector → Second detetor.

3

Hi ChrisDing,

Can Deepstream 4.0 will support more than two detectors engine?

4

Does the direction lib run on CVA/PVA on AGX?
https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/tree/master/anomaly/plugins/gst-dsdirection/dsdirection_lib

5

Back-to-back detecton sample in https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps