Hi, I want to use DeepStream on the TX2 for handling video, running inference on that video, and outputting the video to Amazon Web Services. There are a couple of issues I have though. I want the live video feed to be able to be streamed through AWS, what is the best way to do that? I was thinking an RTSP stream would be best, but it doesn’t seem that DeepStream supports RTSP output. Also, AWS has Greengrass (https://aws.amazon.com/greengrass/) which can run on the TX2, including inference, and interfaces directly with AWS. I would imagine DeepStream would be best for handling the video feed and running inference due to it all being in NVIDIA’s ecosystem, but DeepStream has no mention of interface with AWS, so it seems logical to include Greengrass at some point. I don’t know at what point would be the best place to have the interface between Greengrass and DeepStream. I would imagine somehow sending the composite video from DeepStream into Greengrass would be optimal, but I would love to hear other people’s perspective on this and how to accomplish it.
RTSP support for streaming out is coming in a future release of DeepStream. However, until then you can use one of the one source versions.
Your application can use DeepStream to generate metadata that can be sent thru AWS services. We will look into messaging protocols and will keep the forums updated on any developments.