Recherche avancée

Sur d’autres sites (9129)

• libavfilter : Add dehaze-filter option in existing derain.

  22 août 2019, par Xuewei Meng

  libavfilter : Add dehaze-filter option in existing derain.
  Add the support of dehaze filter in existing derain filter source
  
  code. As the processing procedure in FFmpeg is the same for current
  
  derain and dehaze, we reuse the derain filter source code. The
  
  model training and generation scripts are in repo
  
  https://github.com/XueweiMeng/derain_filter.git
  Reviewed-by : Steven Liu <lq@onvideo.cn>
  
  Signed-off-by : Xuewei Meng <xwmeng96@gmail.com>
  

  • [DH] doc/filters.texi
  • [DH] libavfilter/vf_derain.c

• Why does ffmpeg output have to be encoded entirely before it can be played ?

  17 septembre 2019, par Eves

  I have an MVC C# application (sort of) that calls an api to get a FileStreamResult from an mp4. At the moment the mp4 is your standard h.264 encoded file and we just create a FileStream of the mp4 file and return a new FileStreamResult of that FileStream. Works fine. The mp4 is immediately available in the html video control. However, I’m looking into whether or not we can work with h.265 encoded files. So I am using ffmpeg to translate the file. Obviously there would be a delay due to the use of libx264 to change the h.265 file to h.264. But the real problem I have is that page just sits there until the entire mp4 is transcoded to h.264 before any part is viewable.

  API class variable :

  private Process _videoConcatenationProcess;

  API code :

  if (_videoConcatenationProcess != null && !_videoConcatenationProcess.HasExited)
  
  {
  
      _videoConcatenationProcess.Kill();
  
      _videoConcatenationProcess.Dispose();
  
  }
  
  _videoConcatenationProcess = new Process();
  
  _videoConcatenationProcess.StartInfo.UseShellExecute = false;
  
  _videoConcatenationProcess.StartInfo.RedirectStandardOutput = true;
  
  _videoConcatenationProcess.StartInfo.RedirectStandardError = true;
  
  _videoConcatenationProcess.StartInfo.CreateNoWindow = true;
  
  _videoConcatenationProcess.EnableRaisingEvents = true;
  
  _videoConcatenationProcess.StartInfo.FileName = @"C:\ffmpeg.exe";
  
  _videoConcatenationProcess.OutputDataReceived += (o, k) => System.Diagnostics.Debug.WriteLine(k.Data ?? "NULL", "ffplay");
  
  _videoConcatenationProcess.ErrorDataReceived += (o, k) => System.Diagnostics.Debug.WriteLine(k.Data ?? "NULL", "ffplay");
  
  _videoConcatenationProcess.Exited += (o, k) => System.Diagnostics.Debug.WriteLine("Exited", "ffplay");
  
  _videoConcatenationProcess.StartInfo.Arguments = $" -i \"{videoPath}\" -c:v libx264 -preset ultrafast -maxrate 3000k -bufsize 12000k -pix_fmt yuv420p -g 50 -vf scale=iw/2:-2 -c:a copy -movflags +faststart+frag_keyframe -f mp4 -";
  
  _videoConcatenationProcess.Start();
  
  _videoConcatenationProcess.BeginErrorReadLine();
  
  return new FileStreamResult(_videoConcatenationProcess.StandardOutput.BaseStream, mimeType);

  Simple html in the cshtml for the video :

  <video controls="controls" preload="metadata" controlslist="nodownload">
  
    <source src="@Html.Raw( Model.VideoUrl )" type="@Model.MimeType">
  
    Your browser does not support HTML5 video.
  
  </source></video>

  I’ve been tweaking the ffmpeg parameters for speed/quality and I am not done with that. However, the real problem is that the page starts up and the video control just sits there with the spinning thing until the entire mp4 file is encoding with h.264. I am using "-movflags +faststart+frag_keyframe" and the video control has preload="metadata" so shouldn’t that have allowed the video to be available at least in part early on ?

  So what I am I missing that the video control cannot show the video until it is fully encoded ?

  Here is the console output showing that the metadata is generated before the video has been converted :

  ffplay: Output #0, mp4, to 'pipe:':
  
  ffplay:   Metadata:
  
  ffplay:     major_brand     : isom
  
  ffplay:     minor_version   : 512
  
  ffplay:     compatible_brands: isomiso2mp41
  
  ffplay:     encoder         : Lavf58.20.100
  
  ffplay:     Stream #0:0(und): Video: h264 (libx264) (avc1 / 0x31637661), yuv420p, 960x540, q=-1--1, 30 fps, 15360 tbn, 30 tbc (default)
  
  ffplay:     Metadata:
  
  ffplay:       handler_name    : VideoHandler
  
  ffplay:       encoder         : Lavc58.35.100 libx264
  
  ffplay:     Side data:
  
  ffplay:       cpb: bitrate max/min/avg: 1000000/0/0 buffer size: 2000000 vbv_delay: -1
  
  ffplay:     Stream #0:1(und): Audio: aac (LC) (mp4a / 0x6134706D), 44100 Hz, stereo, fltp, 95 kb/s (default)
  
  ffplay:     Metadata:
  
  ffplay:       handler_name    : SoundHandler
  
  ffplay: frame=    8 fps=0.0 q=24.0 size=       0kB time=00:00:00.39 bitrate=   0.7kbits/s speed=0.565x    
  
  ffplay: frame=   24 fps= 19 q=36.0 size=       0kB time=00:00:00.95 bitrate=   0.3kbits/s speed=0.766x    
  
  ...

• Is a any way of piping input to deepspeech under Linux ?

  24 octobre 2019, par Unreal Admin

  I am trying to continually stream audio from my IP camera to a server running deepspeech to decode the audio stream to text in realtime using FFMPEG.

  I am using the following command :

  $ffmpeg -i rtsp ://192.168.1.249:8080/h264_pcm.sdp -ar 16000 -ac 1 -acodec pcm_s16le -vn -f wav - |
  deepspeech —model deepspeech-0.5.1-models/output_graph.pbmm —alphabet deepspeech-0.5.1-models/alphabet.txt —lm deepspeech-0.5.1-models/lm.binary —trie deepspeech-0.5.1-models/trie —audio -

  Deepspeech : FileNotFoundError : [Errno 2] No such file or directory : ’-’

  deepspeech doesn’t like the ’—audio -’ specifier.

  Is there any way of achieving the above goal using this approach ? Perhaps deepspeech just isn’t designed to do this ?

  Are there any other suggestions as to how this might be achieved ?

  ffmpeg can be directed to write a .wav file. Deepspeech and transcode this file. But I wish to continually stream and transcode the audio.