I'm working on a video streaming wearable device. During the tests, it came up that the pipeline clock and stream stop while fast walking or running. It's bizarre behaviour because in debug messages there are no errors about the broken pipeline, besides lost frames. It's frizzed and only restarting help. May you guys guess what causes the problem?
The pipelines I use:
streaming device:
gst-launch-1.0 -vem --gst-debug=3 v4l2src device=/dev/video0 ! video/x-raw,width=640,height=480,framerate=\(fraction\)30/1 ! v4l2h264enc extra-controls=s,video_bitrate=250000 capture-io-mode=4 output-io-mode=4 ! "video/x-h264,level=(string)4" ! rtph264pay config-interval=1 ! multiudpsink clients="127.0.0.1:5008,10.123.0.2:5008"
client:
udpsrc port=5008 do-timestamp=true ! application/x-rtp, media=(string)video, clock-rate=(int)90000, encoding-name=(string)H264, payload=(int)96 ! rtpjitterbuffer latency=100 drop-on-latency=true drop-messages-interval=100000000 ! queue max-size-buffers=20000 ! rtph264depay ! h264parse ! avdec_h264 ! videoconvert ! glupload ! qmlglsink name=qmlglsink sync=false
The hardware I use is a PS3 Eye cam, and LTE modem to transmit video with a pretty low uplink of 1-2 Mbit/s, and everything running on RaspberryPi 3b+ 1GB.
For more debug info there are also pictures of the log file after last registered dropped frame and every next "cycle" sends a new query, loops over GST Element from sink to the source which is my camera and ends with max query duration(highlighted query to v4l2src)
Do you know how to overcome this problem?
The problem has been resolved. The issue was not variable encoder bitrate. A more detailed inspection and pipeline that works for me is in this GStreamer issue page
