Recherche avancée

Sur d’autres sites (8284)

• Catching ffmpeg errors in Tensorflow

  12 avril 2018, par lollercoaster

  Currently I make extensive use of tensorflow’s ffmpeg interface :

  from tensorflow.contrib.framework.python.ops import audio_ops as contrib_audio
  
  from tensorflow.python.ops import io_ops
  
  from tensorflow.contrib import ffmpeg
  
  
  
  def load(mp3_path):
  
      with tf.name_scope("loading") as scope:
  
          audio_binary = tf.read_file(mp3_path)
  
          waveform = tf.reshape(ffmpeg.decode_audio(
  
              audio_binary, file_format='mp3', 
  
              samples_per_second=44100, channel_count=1), [-1])
  
          return waveform

  This is wonderful, but sometimes (well, a lot) I get errors like this when looping through a bunch of files to create TFRecord files :

  [mp3 @ 0xdd4400] Header missing
  
  Error while decoding stream #0:0: Invalid data found when processing input
  
  [mp3 @ 0xf2e440] Header missing
  
  Error while decoding stream #0:0: Invalid data found when processing input
  
  [mp3 @ 0x202bf40] invalid new backstep -1
  
  [mp3 @ 0x15c5440] Header missing
  
  Error while decoding stream #0:0: Invalid data found when processing input
  
  [mp3 @ 0x1752460] Header missing
  
  Error while decoding stream #0:0: Invalid data found when processing input
  
  [mp3 @ 0x1f10160] Header missing
  
  Error while decoding stream #0:0: Invalid data found when processing input
  
  [mp3 @ 0x1a16dc0] Header missing
  
  Error while decoding stream #0:0: Invalid data found when processing input
  
  [mp3 @ 0x9f2c80] Header missing
  
  Error while decoding stream #0:0: Invalid data found when processing input

  which shows up in stdout while running this Python script.

  I’d like to catch these somehow ! And prevent the rest of my loop from continuing if the data I’m reading is invalid.

  Currently, these errors just happen and the script happily keeps running, which is probably not the way it should work.

  I’m not making use of the Dataset API for this, so these direct session.run(waveform_op) calls should throw and Exception or something when parsing fails !

  Anyone know how I can do this ?

• Async Javascript function returns undefined

  14 avril 2018, par Leon Wright

  I am trying to run this function but the it keeps returning undefined when I explicitly hardcode the return value.

  const splitVideo = async (sid, video, part) => {
  
      let framesLocation =`${process.cwd()}/${dirs.videoFrames}/${sid}_${part}`;
  
  
  
      console.log(fs.existsSync(framesLocation));
  
  
  
      if(!fs.existsSync(framesLocation)) {
  
          console.log("making dir");
  
          f.s.mkdirSync(framesLocation);
  
      }
  
  
  
      ffmpeg(video)
  
          .on('end', () => {
  
               return "done";
  
           })
  
           .on('error', (err) => {
  
               throw err;
  
           })
  
           .screenshots({
  
                timestamps: [1,2],
  
                filename: `${sid}_${part}/frame-%s.png`,
  
                folder: `${process.cwd()}/${dirs.videoFrames}`
  
           });
  
  };

  Please help this is very frustrating.

• How to fetch both live video frame and timestamp from ffmpeg to python on Windows

  8 mai 2018, par vijiboy

  Searching for an alternative as OpenCV would not provide timestamps for live camera stream (on Windows), which are required in my computer vision algorithm, I found ffmpeg and this excellent article https://zulko.github.io/blog/2013/09/27/read-and-write-video-frames-in-python-using-ffmpeg/
  The solution uses ffmpeg, accessing its standard output (stdout) stream. I extended it to read the standard error (stderr) stream as well.

  Working up the python code on windows, while I received the video frames from ffmpeg stdout, but the stderr freezes after delivering the showinfo videofilter details (timestamp) for first frame.

  I recollected seeing on ffmpeg forum somewhere that the video filters like showinfo are bypassed when redirected. Is this why the following code does not work as expected ?

  Expected : It should write video frames to disk as well as print timestamp details.
  Actual : It writes video files but does not get the timestamp (showinfo) details.

  Here’s the code I tried :

  import subprocess as sp
  
  import numpy
  
  import cv2
  
  
  
  command = [ 'ffmpeg', 
  
              '-i', 'e:\sample.wmv',
  
              '-pix_fmt', 'rgb24',
  
              '-vcodec', 'rawvideo',
  
              '-vf', 'showinfo', # video filter - showinfo will provide frame timestamps
  
              '-an','-sn', #-an, -sn disables audio and sub-title processing respectively
  
              '-f', 'image2pipe', '-'] # we need to output to a pipe
  
  
  
  pipe = sp.Popen(command, stdout = sp.PIPE, stderr = sp.PIPE) # TODO someone on ffmpeg forum said video filters (e.g. showinfo) are bypassed when stdout is redirected to pipes??? 
  
  
  
  for i in range(10):
  
      raw_image = pipe.stdout.read(1280*720*3)
  
      img_info = pipe.stderr.read(244) # 244 characters is the current output of showinfo video filter
  
      print "showinfo output", img_info
  
      image1 =  numpy.fromstring(raw_image, dtype='uint8')
  
      image2 = image1.reshape((720,1280,3))  
  
  
  
      # write video frame to file just to verify
  
      videoFrameName = 'Video_Frame{0}.png'.format(i)
  
      cv2.imwrite(videoFrameName,image2)
  
  
  
      # throw away the data in the pipe's buffer.
  
      pipe.stdout.flush()
  
      pipe.stderr.flush()

  So how to still get the frame timestamps from ffmpeg into python code so that it can be used in my computer vision algorithm...