Recherche avancée

Sur d’autres sites (8515)

• Opencv/FFMPEG retrieving wrong fps while streaming cameras

  15 avril 2016, par Romanzo Criminale

  I’ve have been trying to retrieve the fps information from different camera stream via RTSP. This is done using the OpenCV library which uses the libavormat provided by ffmpeg. However it seems I can’t rely on the information returned by libav depending on the codec or the camera used. I would like to get a better understanding about how this frame rate are computed etc...

  OpenCV is trying to retrieve the fps of the stream using different properties of AVStream, r_frame_rate, avg_frame_rate and time_base.

  double CvCapture_FFMPEG::get_fps() const
  
  {
  
   #if LIBAVCODEC_BUILD >= CALC_FFMPEG_VERSION(54, 1, 0)
  
    double fps = r2d(ic->streams[video_stream]->avg_frame_rate);
  
   #else
  
    double fps = r2d(ic->streams[video_stream]->r_frame_rate);
  
   #endif
  
  
  
   #if LIBAVFORMAT_BUILD >= CALC_FFMPEG_VERSION(52, 111, 0)
  
    if (fps < eps_zero)
  
    {
  
      fps = r2d(ic->streams[video_stream]->avg_frame_rate);
  
    }
  
   #endif
  
  
  
   if (fps < eps_zero)
  
   {
  
    fps = 1.0 / r2d(ic->streams[video_stream]->codec->time_base);
  
   }
  
  
  
   return fps;
  
  }

  I tried two different camera brands and two different codecs H264, and JPEG. Here are the results, I get :

  Camera 1 running at 15fps (default value)

  camera 1, jpeg

  • r_frame_rate = 15/1
  • avg_frame_rate = 0/0
  • time_base = 1/ 90000
  • tick_per_frame = 1

  camera 1, h264

  • r_frame_rate = 15/1

  • avg_frame_rate = 15/2

  • time_base = 1/ 90000

  • tick_per_frame = 2

  Camera 2 running at 25fps default but the fps is changed within the rtsp url to 12

  camera 2, jpeg

  • r_frame_rate = 12/1

  • avg_frame_rate = 0/0

  • time_base = 1/ 90000

  • tick_per_frame = 1

  camera 2, h264 running at 12fps

  • r_frame_rate = 25/1

  • avg_frame_rate = 0/0

  • time_base = 1/ 90000

  • tick_per_frame = 2

  Also i tried streaming using ffmpeg command line

  ffmpeg -i rtsp://url

  and it seems that only jpeg can return the expected fps.

  Here are the questions have been asking myself :

  1/ what’s really r_frame_rate ?
  It says in the api "[i]it is the smallest frame rate that can accurately represent all timestamps. For example, if you have mixed 25 and 30 fps content, then
  will be 150 (it is the least common multiple).[/i]" I doesn’t make sense to me that it the same time the smallest frame rate (25) and the least common multiple (150).

  2/ How is the avg_frame_rate computed ?
  I had a look at the code but couldn’t really figure out how is this average computed. It seems that most of the time this is returning 0.

  3/ Is there a bug getting the frame rate while using h264 ?
  Using jpeg, it looks like r_frame_RATE can return the actual fps value, but h264 will not work. Are those information not available using h264 ?

  Any help would be appreciated, I spent two days trying to understand why i wasn’t getting the correct values but at this point i’m pretty stuck.
  Thanks in advance !

• vc2enc : carry over quantization index across frames as a starting point

  27 février 2016, par Rostislav Pehlivanov

  vc2enc : carry over quantization index across frames as a starting point
  Previously a global average was used. Using the previous quantizer
  
  resulted in a fairly significant speedup as slice size selection settled
  
  down quicker.
  Signed-off-by : Rostislav Pehlivanov <atomnuker@gmail.com>
  

  • [DH] libavcodec/vc2enc.c

• Storing large videos on the Server

  14 novembre 2019, par civ15

  In my application the client is uploading relatively large videos, average is 500 MB. I have to encode the video in the backend to reduce the size, and I’m trying to use ffmpeg for that. However, ffmpeg needs around 10 minutes for each video, and I should be able to serve the video to the client instantly, or, at least in a few seconds after the upload is finished.

  What would be the best practice in this context ?