In the past I have successfully deployed a DS project running on a Jetson Nano and doing object detection with up to three USB cameras in parallel.
For a new project this is the requirement:
- The video is delivered via WebRTC to a Kurento Media Server in the cloud, preferably but not necessarily AWS (Ubuntu based, x86_64)
- In the Media Server there an “OpenCV” plugin (C++), which provides a way to intercept the incoming video RGB frame by RGB frame, so that it can be processed as an OpenCV::Mat.
- I’m wondering now, if I could insert some “AI” code utilizing the DS SDK, processing each frame and doing the same as I did before with the edge app, of course also with GPU support by the cloud.
Would that generally be possible? Could you provide me some starters, if yes?