• Hardware Platform (Jetson / GPU)
Jetson AGX Xavier
• DeepStream Version
6.3
• JetPack Version (valid for Jetson only)
5.1.4
• Question
I would like to send the inference results of a deepstream application downstream to another application running on the same device that executes some business logic.
For example, the deestream app detects the number of vehicles in each frame, while the downstream application calculates the average number of vehicles seen per minute.
I am currently using a probe function that sends a websocket message with the inference results to the downstream process. Is there a way to do this more efficiently with nvmsgbroker plugin?
I am under the impression the supported protocols such as AMQP/Kafka/MQTT are more appropriate for device → cloud than process → process and have larger overheads than simple socket/HTTP/TCP messages. If not, what would be the best approach for this?