Recherche avancée

Sur d’autres sites (6613)

• Revision 45367 : [source:_plugins_/step Step]. Deux déclarations conditionnelles pour ...

  13 mars 2011, par esj@… — Log

  Step. Deux déclarations conditionnelles pour pouvoir essayer Step en SPIP 2.2.

• Revision 45367 : [source:_plugins_/step Step]. Deux déclarations conditionnelles pour ...

  13 mars 2011, par esj@… — Log

  Step. Deux déclarations conditionnelles pour pouvoir essayer Step en SPIP 2.2.

• How to I increase IO buffer size in Crystal ?

  28 avril 2017, par timur

  I have an IO class that will send data it receives across all connected web sockets. The data is received from a Process call running an FFMPEG command. The class is as so :

  class FfmpegBinaryIO
  
      include IO
  
  
  
      def initialize(@sockets = [] of HTTP::WebSocket)
  
      end
  
  
  
      def read(slice : Bytes)
  
          raise "FfmpegBinaryIO does not support reads"
  
      end
  
  
  
      def write(slice : Bytes)
  
          @sockets.each { |socket| socket.send Base64.encode(slice) }
  
      end
  
  end

  The write function will iterate the array of sockets I pass and send the chunks encoded in base64. The chunks are received by using the following Process call where binary_output is an instance of the FfmpegBinaryIO class :

  spawn do
  
      Process.run(command: ffmpeg_command, output: binary_output, error: ffmpeg, shell: true)
  
  end

  The class is working as it should and I am able to display the chunks using the following Javascript code :

  const img = document.getElementById('view')
  
  const ws = new WebSocket('ws://127.0.0.1:8080/ffmpeg')
  
  ws.onmessage = msg => {
  
      console.dir(msg)
  
      view.src = `data:image/jpeg;base64,${msg.data}`
  
  }

  However, the frames appear to be cut in half when displayed and it skips every other frames, presumably because the data is corrupted. What I found out when writing the same Bytes to JPG files is that chunks would be incomplete and as such every other file written was corrupted.

  This is not the kind of behavior I experienced when using a similar approach in other programming languages. My hypothesis is that chunks are being cut due a buffer size limit imposed. Is there any way I can have every write call represent a single frame chunk from FFMPEG by increasing this hypothetical buffer size ?

  UPDATE

  My suspicions are right that this has to do with the size of slices that the FfmpegBinaryIO receives. I tried modifying the FFMPEG command to reduce the output byte size by reducing the scaling as so :

  ffmpeg_command = "ffmpeg -i /dev/video0 -framerate 6 -vf scale=320:240 -r 24 -qscale:v 35 -updatefirst 1 -f image2 -y pipe:1"

  Where previously the command I used did not include the -vf scale=320:240 option.