search

motioneye-project / motioneye

A web frontend for the motion daemon.
GNU General Public License v3.0
3.99k stars 655 forks source link

YUV420P issue with decoder = vaapi #3015

Open mdPlusPlus opened 4 months ago

mdPlusPlus commented 4 months ago

I'm using following lines in my camera.conf to use hardware decoding (VA-API):

...
netcam_url rtsp://mycamera.lan:12345/stream
netcam_params decoder = vaapi,keepalive = off,tolerant_check = on,rtsp_transport = tcp
...

If I check my motion.log, I find the following hint:

netcam_rtsp_ntc: ******************************************************
netcam_rtsp_ntc: The image sent is being 
netcam_rtsp_ntc: trancoded to YUV420P.  If possible change netcam 
netcam_rtsp_ntc: picture format to YUV420P to possibly lower CPU usage.
netcam_rtsp_ntc: ******************************************************

And I notice a small CPU utilization increase, which makes me think the format transformation is done on CPU, before being sent to GPU.

However, when I check the actual pixel format of the stream it is YUV420P:

$ ffprobe -hide_banner rtsp://mycamera.lan:12345/stream

[rtsp @ 0x55e2cb8ad700] max delay reached. need to consume packet
[rtsp @ 0x55e2cb8ad700] RTP: missed 1 packets
Input #0, rtsp, from 'rtsp://mycamera.lan:12345/stream':
  Metadata:
    title           :  
  Duration: N/A, start: 0.162811, bitrate: N/A
  Stream #0:0: Video: h264 (High), yuv420p(progressive), 1920x1080, 30 fps, 30 tbr, 90k tbn

Is this somehow user error or is there something wrong with the detection of the used pixel format?
I see there is an input_format option for netcam_params, but from the documentation it does not seem related.