Recherche avancée

Sur d’autres sites (6185)

• ffmpeg simple 5fps webcam capturing but results in poor performance

  16 septembre 2022, par irous

  I'm using Windows 10, intel core i5, 8GB ram. and download ffmpeg from official site : https://www.gyan.dev/ffmpeg/builds/

  


  I'm trying to capture 5fps video from webcam but the output fps is only 3.4 and speed is 0.668x. (I also tried with libx264 but the speed is same 0.5).
  
This is my command :

  


  ffmpeg ^
  -f dshow -r 5 -s 640x480 -i video="Sony Visual Communication Camera" ^
  -f rawvideo output.raw


  


  screenshot of console output :


  


  The cpu usage for above ffmpeg command is about 0.1% and also ram/disk usage is relatively low.
  
If I change the fps to 30 (-r 30), the output is : fps 7.9, speed 0.261x and CPU usage is still very low 0.1%.

  


  But I can easily record 30fps video with the same format yuyv422 in Directshow GraphEdit :


  


  So my question is why does ffmpeg show such poor performance even with 5fps video capturing ? If it can only produce 3.4 fps in the first case why it can output up to 7.9 fps in the latter case (when I set desired fps to 30) ? Did I mis-config ffmpeg ?
  
Any help is appreciated.

  


• Build PJSIP with video support on Windows 7, with MinGW

  3 juillet 2015, par Daniel V

  I’m trying to build PJSIP with video support for Windows OS, on my Windows 7 PC by using MinGW.
  Following the official guide from PJSIP :
  http://trac.pjsip.org/repos/wiki/Getting-Started/Autoconf#VideoSupportfor2.0andabove

  Building PJSIP without video support works as expected for me.

  • I’m using the latest PJSIP 2.2.1 from SVN
  • SDL2-devel-2.0.3-mingw.tar.gz (MinGW 32/64-bit)
  • ffmpeg-20140805-git-de41798-win32-dev

  I have added "#define PJMEDIA_HAS_VIDEO 1" in the config_site.h file
  and I’m building PJSIP with the following options :
  ./configure —with-ffmpeg="/c/PJSIP/ffmpeg" —with-sdl="/c/PJSIP/SDL"

  but I have the following compilation error for SDL :

  c:/PJSIP/SDL/lib/libSDL2main.a(SDL_windows_main.o): In function `console_main':
  
  /Users/slouken/release/SDL/SDL2-2.0.3-source/foo-x86/../src/main/windows/SDL_win
  
  dows_main.c:140: undefined reference to `SDL_main'
  
  collect2.exe: error: ld returned 1 exit status
  
  make[2]: *** [../bin/pjsua2-test-i686-pc-mingw32] Error 1
  
  make[2]: Leaving directory `/c/PJSIP/trunk/pjsip/build'
  
  make[1]: *** [pjsua2-test-i686-pc-mingw32] Error 2
  
  make[1]: Leaving directory `/c/PJSIP/trunk/pjsip/build'
  
  make: *** [all] Error 1

  The same error is available with SDL-2.0.2 too.

• How to pack the pyav.packet and distribute to another computer

  31 juillet 2024, par lambertk

  I'm currently working on projects which needs to read frames from RTSP server on single entry of computer, do some preprocessing and distribute these frames with preprocessed metadata to different backend for different purpose.

  


  And after googling, I found that PyAV could be the solution which can retrieve the video from RTSP source and make it packets, which could possibly be sent to another computer.

  


  Considering the network bandwidth, transmit the packets instead of the decoded frames could be better choice.

  


  But now comes the problem, socket/MQ, usually only allows to send bytes or string.
  
Encode the PyAV.packet.Packet object into byte is easy by bytes(packet), but I couldn't find out the way to decode it back to PyAV.packet.Packet object.

  


  I've tried to use pickle to serialize the packet, but this method is not implemented in PyAV, and was rejected by the official team.

  


  I've also tried to use another package called msgpack, which also failed to serialize the packet.

  


  I've tried the following code after reading the source code of PyAV

  


  packet_bytes = bytes(packet)
pt = av.packet.Packet(len(packet_bytes))
pt.update(packet_bytes)


  


  the update function seems did not update anything

  


  Is there anyway to decode the bytes back to packet object ?

  


  Or, can someone give out a way to encode the frame packet and the preprocessed metadata (which is differ frame by frame) together (like H264 SEI Message, which I tried, but could not be inserted when using Python) then send to backend ?