This is more of a conceptual question.
I was trying to grab meta_data from a probe and use it somewhere else. The more logic implemented in the probe lead to a slower videofeed. It also had scale problems because the more metadata outputted per object led to slower performance.
How does the message-conv and message-broker plugins successfully extract metadata without slowing down the pipeline when more objects appear on the screen?
Implementing compute functionality inside probe is not advisible as it is a blocking call. Message broker and converter are separate gstreamer plugins and run in different space in parallel. For better performance you should implement a custom gstreamer plugin to achieve required functionality.
Thank you for the response. I suspected the probe was a bad idea.
Last question, how does the message broker gstreamer plugin interact with for example rabbitmq plugin?
Does this plugin just prep the metadata for a third party plugin to take?
message broker is one broker plugin which can interact with protocol adapter, like amqp, kafka, azure, and customized adapter, through interface, nvds_msgapi
details you can see here, https://docs.nvidia.com/metropolis/deepstream/plugin-manual/index.html#page/DeepStream%20Plugins%20Development%20Guide/deepstream_plugin_details.3.14.html#wwpID0E0BQ0HA
This plugin sends payload messages to the server using a specified communication protocol. It accepts any buffer that has NvDsPayload metadata attached and uses the nvds_msgapi_* interface to send the messages to the server.
Thank you again. All questions are answered. Please lock the thread.