actfw-gstreamer

actfw's components using GStreamer for implementation. actfw is a framework for Actcast Application written in Python.

Installation

sudo apt-get update
sudo apt-get install -y python3-pip python3-pil 
sudo apt-get install libgstreamer1.0-dev libgirepository1.0-dev ibgstreamer-plugins-base1.0-dev libglib2.0-dev gcc
pip3 install actfw-gstreamer

Document

• API References

Usage

See actfw-core for basic usage of actfw framework.

Initalization

An application using actfw-gstreamer have to initialize GStreamer library before using actfw-gstreamer's components.

if __name__ == '__main__':
    import gi

    gi.require_version('Gst', '1.0')
    from gi.repository import Gst
    Gst.init(None)

    main()

videotestsrc

You can learn basic usage of actfw-gstreamer by using videotestsrc.

from actfw_gstreamer.capture import GstreamerCapture
from actfw_gstreamer.gstreamer.converter import ConverterPIL
from actfw_gstreamer.gstreamer.stream import GstStreamBuilder
from actfw_gstreamer.restart_handler import SimpleRestartHandler


def videotestsrc_capture() -> GstreamerCapture:
    pipeline_generator = preconfigured_pipeline.videotestsrc()
    builder = GstStreamBuilder(pipeline_generator, ConverterPIL())
    restart_handler = SimpleRestartHandler(10, 5)
    return GstreamerCapture(builder, restart_handler)


def main():
    app = actfw_core.Application()

    capture = videotestsrc_capture()
    app.register_task(capture)

    consumer = YourConsumer()
    app.register_task(consumer)

    capture.connect(consumer)

    app.run()

This generates Frames using videotestsrc.

• GstreamerCapture is a Producer.
  • It generates Frames consists of an output of ConverterBase. In this case, converter class is ConverterPIL and output is PIL.Image.Image.
• GstStreamBuilder and PipelineGenerator determines how to build gstreamer pipelines.
• preconfigured_pipeline provides preconfigured PipelineGenerators.
• SimpleRestartHandler is a simple implementation of RestartHandlerBase, which determines "restart strategy".

For more details, see tests.

rtspsrc

You can use rtspsrc using preconfigured_pipeline.rtsp_h264().

Note that, as of now (2021-04), Actcast application cannot use multicast UDP with dynamic address and unicast UDP. (RTSP client communicates with RTSP server in RTP and determines adderss of mulitcast UDP.) Therefore, you can use only the option protocols = "tcp". See also https://gstreamer.freedesktop.org/documentation/rtsp/rtspsrc.html#rtspsrc:protocols .

You should also pay attention to decoders. Available decoders are below:

If your application supports various devices, you should branch by hardware types and select appropriate decoder_type. For example, it is recommended to use decoder_type omx for Raspberry Pi 3 and v4l2 for Raspberry Pi 4. Currently, this library does not provide auto determination.

Development Guide

Installation of dev requirements

pip3 install poetry
poetry install

Running tests

poetry run pytest -v

Releasing package & API doc

CI will automatically do. Follow the following branch/tag rules.

1. Make changes for next version in master branch (via pull-requests).
2. Make a PR that updates version in pyproject.toml and merge it to master branch.
3. Create GitHub release from master branch's HEAD.
   1. Draft a new release.
   2. Create new tag named release-<New version> (e.g. release-1.4.0) from Choose a tag pull down menu.
   3. Write title and description.
   4. Publish release.
4. Then CI will build/upload package to PyPI & API doc to GitHub Pages.