Recherche avancée

Sur d’autres sites (4947)

• python opencv + ffmpeg makes cpu more busy

  15 juillet 2022, par shao wenzhi

  I use cv2 to read frames of local video, then use ffmpeg to push the stream to remote rtmp server.

  


  The command is :

  


  command = ['ffmpeg',
           '-y', '-an',
           '-f', 'rawvideo',
           # '-vcodec', 'rawvideo',
           '-pix_fmt', 'bgr24',
           '-s', sizeStr,
           '-r', '25',
           '-i', '-',
           # '-c:v', 'libx264',
           # '-pix_fmt', 'yuv420p',
           '-preset', 'ultrafast',
           '-f', 'flv',
           rtmp]


  


  I found that compared with pushing the video directly through ffmpeg, the CPU occupation of "cv2 + ffmpeg" is much more higher. which is : ffmpeg : 5% CPU, cv2 + ffmpeg : 18% CPU

  


  One possible cause is that cv2 turns the frame into BGR, and the ffmpeg turns it into RBG back again, because I found when I removed the option '-pix_fmt', 'bgr24', the CPU occupation drops into 5%, but the video pulled from server turns out to be strange, like this :

  


  

  


  Is there any method to reduce the cost of converting BGR into RGB ??

  


• Anomalie #4707 (Nouveau) : SPIP n’indique plus qu’une nouvelle version disponible

  29 mars 2021

  La dernière version SPIP 3.2.10 est sortie mais on ne le signale pas dans le backoffice

  si on lance le genie de mise à jour sur un SPIP 3.2.7 ou SPIP 3.2.9 par ex. on ne trouve rien.

  pourtant le fichier archives.xml semble bien renseignée avec archives/spip-v3.2.11.zip
  https://files.spip.net/spip/archives.xml

  en regardant le code, cela a l’air de coincer au niveau de la regex
  https://git.spip.net/spip/spip/src/branch/3.2/ecrire/genie/mise_a_jour.php#L123

  je ne sais pas s’il y a qqchose de mal formaté ou non.

• Media player get stuck in the middle of a buffered range on Chrome

  29 septembre 2019, par Feng Yu

  WHAT IS MY PROBLEM ?

  My website’s live streaming player use hls.js. From my server’s stat, there is many case where player get stuck in the middle of a buffered range.

  Here is my server raw stat log(removed some useless params) :

  tm=2019-09-27 12:04:41`bufferLevel=8.447303999999974`currentTime=158.4`buffered=[6.024,166.832]`readyState=4`ua=Mozilla/5.0 (Windows NT 6.2; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/53.0.2785.116 Safari/537.36 QBCore/3.53.1153.400 QQBrowser/9.0.2524.400 Tencent AppMarket/4.8 GameCenter

  currentTime is got by HTMLMediaElement.currentTime and buffered is got by HTMLMediaElement.buffered :

  currentTime=158.4
  
  buffered=[6.024,166.832]
  
  readyState=4

  From W3c :

  If HTMLMediaElement.buffered contains a TimeRange that includes the current playback position and enough data to ensure uninterrupted playback :

  1. Set the HTMLMediaElement.readyState attribute to HAVE_ENOUGH_DATA.
  2. Playback may resume at this point if it was previously suspended by a transition to HAVE_CURRENT_DATA.

  In this case, 613.3 is in the middle of [469.277,677.612], video should be progressing, but it is not.

  Hls.js will periodly check currentTime has progressed every 100ms. if currentTime has not progressed for 1000ms, then hls.js will trigger STALL event and I will send a stall stat to server.

  I cannot reproduce this problem on my side, it only appears on my server stat.

  WHAT I’VE TRIED

  shaka player has a module detect this case(https://www.ellealcatrase.eu/player2/docs/api/lib_media_stall_detector.js.html), Its comment shows that :

  Some platforms/browsers can get stuck in the middle of a
  buffered range (e.g. when seeking in a background tab). Detect when
  we get stuck so that the player can respond.

  but I cannot reproduce when my browser is in a background tab.