Note beforehand: The code can be found
here.
I'm using GStreamer to transmit data between Docker containers to use in Machine Learning algorithms. I am using GStreamer 1.0 with the Python bindings. My Python version is 3.6
I'm am trying to implement the following pipeline:
filesrc ! decodebin ! jpegenc ! rtpjpegpay ! udpsink
I've tried this pipeline and it works (also in the Docker container). Since I need to dynamically set the host for the udpsink and the file location for the filesrc I'm assembling the pipeline manually using Python.
The steps I've taken following the Dynamic Pipeline tutorial:
1. Initialize GStreamer
2. Create all the elements using Gst.ElementFactory.make()
3. Create a pipeline and add all elements using Gst.Pipeline.new() and pipeline.add()
4. Set properties of filesrc and udpsink
5. Link filesrc to decodebin, jpegenc to rtpjpegpay and rtpjpegpay to udpsink.
6. Add a handler for the 'pad-added' signal for the decodebin
7. set the state of the pipeline to PLAYING
8. Try to link the elements when the handler is called.
The handler is called a first time when the decodebin creates a source pad 'src0' which has the capability 'audio/x-raw'. Since I'm interested in video, I ignore this pad, and wait for the video pad. Unfortunately, the handler is never called again and the pipeline stalls.
I don't know whether or not the handler is called again or not. I also don't know if the pad actually get's created or not.
My questions are:
1. Why is my handler only called once? Is this because of the handler itself or is it the pad which isn't created?
2. How do I get debug feed from GStreamer to Python's stdout stream? As for now, I haven't found how to redirect it or change the level, which gives me limited feedback on my attempts to get the stream to run.
I've posted the full gist
here.
Many thanks!
Brecht
--
Brecht Verhoeve
Master Student Computer Science Engineering
Ghent University
Locked
