Azure Video Analyzer on Jetson Nano with GPU support

Question

Hi,

Is there a way to use the Azure Video Analyzer archtitecture on Nvidia Jetson Edge devices (Nano or Xavier NX) with a GPU accelerated inference server?

So far I have only found tutorials using Nvidias Deepstream SDK as a substitute for the Azure Video Analyzer (AVA) module. However, by using the Deepstream SDK I wouldn't be able to take advantage of AVA's pipeline topology architecture for quick deployment of different use cases.

Instead, I would like to deploy my own inference server as a grpc extension module within the AVA pipeline (as indicated here). This would ideally be based on a container that sends inference jobs to the jetson gpu (maybe Nvidias Triton Server container?).

This seems to be the case in Microsofts OpenVino example, so essentially I would like to replicate this for Nvidias edge devices.

Answer 1

Hi Yutong,
thank you for the reply. I'have submitted a support ticket already, thank you.

It definitely seems possible, but atm there is no official guide how to implement AVA on jetson devices.

Thanks,
Robby.

Answer 2

Please see this for an example of how to build an inference server around DeepStream SDK, such that the server can be used with AVA.