[meta-freescale] img2dcompositor

Cristiano Chieppa cristiano.chieppa at gmail.com
Mon Jul 3 01:06:36 PDT 2017


Dear All, I'm struggling with the imx-gstreamer on a SABRE iMX6 quad core. I have prepare an image with Ubuntu 16 and compiled gstreamer iMX. I would like to use the imxg2dcompositor to gather 4 h264 streams. 
 
The pipeline I have prepared, following some docs on the Network, is the following:
gst-launch-1.0 \
imxg2dcompositor name=c background-color=0xFF \
sink_0::xpos=0 sink_0::ypos=0 sink_0::width=400 sink_0::height=300 \
sink_1::xpos=400 sink_1::ypos=0 sink_1::width=400 sink_1::height=300  \
queue2 ! video/x-raw, width=800, height=600 ! imxipuvideosink \
rtspsrc location=rtsp://<user>:<pwd>@192.168.1.250/h264_2 ! rtph264depay ! h264parse ! imxvpudec ! queue2 ! c.sink_0 \
rtspsrc location=rtsp://<user>:<pwd>@192.168.1.188/h264_2 ! rtph264depay ! h264parse ! imxvpudec ! queue2 ! c.sink_1
I have got to face with the following  problem: the video is black, none of the streams is visible.  

Please note that if i show the 4 streams not grouped, but arranged with x,y coordinates in a sink like img2dvideosink, it works perfectly.

Moreover, if I convert the pipe on my Ubuntu workstation (with the "normal" compositor, and using avdec_h264 + videoconvert ) it works like a charm.

Finally, if I use the same stream (2 times the H264 from a same camera), it works!
 
I have tried several tests to solve the problem, went to any corner of google, but I haven't found out any interesting answer.
 
Btw, I have also tried the imxipucompositor: same problem.
 
Where I wrong?
 
Thank you for any advise. I’m going crazy.


PS: I have tried to use on the imx pipeline above the software “compositor”: it works and I can see the 4 streams, but is not acceptable because is too slow. So, I come back to my question: why it doesn’t work in this way?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.yoctoproject.org/pipermail/meta-freescale/attachments/20170703/1973d365/attachment.html>


More information about the meta-freescale mailing list