Migrating Traditional DeepStream Apps to Service Maker Apps in Python

There are two options available for migrating traditional Deepstream applications to Python apps. One option involves using Pipeline APIs with configuration files, while the other option utilizes Flow APIs. Each approach offers its own advantages and can be chosen based on the specific requirements and preferences of the application developers.

Guidelines for migrating Deepstream applications to Python apps using pipeline APIs remain consistent with those of Service Maker C++ APIs. Thus, this section will focus on how to migrate Deepstream Applications to Python apps using Flow APIs.

1. Defining the Video source

   Service Maker uses a Source List YAML file to define inputs, including the source information and input configuration. Specification of source list file can be found in SourceConfig. Once we have a Source List file, we can create a batch capture flow with it:

   from pyservicemaker import Pipeline, Flow
   Flow(Pipeline("my-pipeline")).batch_capture("my_source_list.yaml")

2. Pipeline Customization

   In Flow APIs, each predefined section for pipeline customization in the configuration file for a traditional Deepstream Application corresponds to its respective function:

   DS Configuration	Flow API Name
   primary-gie	infer
   secondary-gie	infer
   tracker	track

3. Sink management

   Flow APIs have corresponding functions for each sink type supported in traditional Deepstream Application:

   DS Sink Type	Flow API Name
   Fakesink	render(mode=RenderMode.DISCARD)
   Display	render(mode=RenderMode.DISPLAY)
   File	encode(“/tmp/sample.mp4”)
   UDP	encode(“udp://localhost:5400”)
   RTSP	encode(“rtsp://localhost:5400”)
   MsgConvBroker	publish

   To enable multiple-sink, we can just call fork() to fork the current flow and append various sink functions with it (check the ‘fork’ function in previous sections).