I am evaluating option of DeepStream VS Gstreamer for capturing video as well as posting binary frame data to Kafka.
I can see Deepstream requires custom component as we required to send entire frame as part of sink. Whereas, Gstreamer its going to be straightforward as we are going to read the frames and post the frames to Kafka before converting the same into binary format.
In terms of performance and resource utilisation, which one do you recommend?
DeepStream SDK is based on gstreamer and for sending data to kafka server, you can refer to deepstream-test4 and deepstream-test5 samples.
However, I am looking to send entire image in binary format to Kafka. Whereas, current component can be used only to send processed json data as part of metadata. Is there possibility that we can use the same Kafka component for sending image frame in binary format? or do you think we need to write our own custom component?
You would need to create user meta to put entire images. There is a sample demonstrating encoding the detected objects to JPGs and putting in user meta:
You may refer to it for your usecase.
Let us try to implement using GStreamer. If that does not work, we will come back and build custom metadata.