Recherche avancée

Sur d’autres sites (6680)

• FFMpeg fails to detect input stream when outputting to pipe's stdout

  27 septembre 2020, par La bla bla

  We have h264 frames as individual files, we read them to a python wrapper and piping them to ffmpeg.

  


  ffmpeg subprocess is launched using

  


      command = ["ffmpeg",
               "-hide_banner",
               "-vcodec", "h264",
               "-i", "pipe:0",
               "-video_size", "5120x3072",
               '-an', '-sn',  # we want to disable audio processing (there is no audio)
               '-pix_fmt', 'bgr24',
               "-vcodec", "rawvideo",
               '-f', 'image2pipe', '-']
    pipe = sp.Popen(command, stdin=sp.PIPE, stdout=sp.PIPE, bufsize=10 ** 8)


  


  Our goal is to use ffmpeg to convert the individual h264 frames into raw BGR data that we can manipulate using OpenCV.

  


  the files are read in a background thread and piped using

  


      ...
    for path in files:
        with open(path, "rb") as f:
            data = f.read()
            pipe.stdin.write(data)


  


  When we try to read the ffmpeg's output pipe using

  


      while True:
        # Capture frame-by-frame
        raw_image = pipe.stdout.read(width * height * 3)


  


  we get

  


  [h264 @ 0x1c31000] Could not find codec parameters for stream 0 (Video: h264, none): unspecified size
Consider increasing the value for the 'analyzeduration' and 'probesize' options
pipe:0: could not find codec parameters
Input #0, h264, from 'pipe:0':
  Duration: N/A, bitrate: N/A
    Stream #0:0: Video: h264, none, 25 tbr, 1200k tbn, 50 tbc
Output #0, image2pipe, to 'pipe:':
Output file #0 does not contain any stream


  


  However, when I change the sp.Popen command to be

  


  
    f = open('ffmpeg_output.log', 'wt')
    pipe = sp.Popen(command, stdin=sp.PIPE, stdout=f, bufsize=10 ** 8) # Note: the stdout is not f


  


  we get the gibberish (i.e, binary data) in the ffmpeg_output.log file, and the console reads

  


  [h264 @ 0xf20000] Stream #0: not enough frames to estimate rate; consider increasing probesize
[h264 @ 0xf20000] decoding for stream 0 failed
Input #0, h264, from 'pipe:0':
  Duration: N/A, bitrate: N/A
    Stream #0:0: Video: h264 (Baseline), yuv420p, 5120x3072, 25 fps, 25 tbr, 1200k tbn, 50 tbc
Output #0, image2pipe, to 'pipe:':
  Metadata:
    encoder         : Lavf56.40.101
    Stream #0:0: Video: rawvideo (BGR[24] / 0x18524742), bgr24, 5120x3072, q=2-31, 200 kb/s, 25 fps, 25 tbn, 25 tbc
    Metadata:
      encoder         : Lavc56.60.100 rawvideo
Stream mapping:
  Stream #0:0 -> #0:0 (h264 (native) -> rawvideo (native))
Invalid UE golomb code
    Last message repeated 89 times
Invalid UE golomb code
    Last message repeated 29 times
Invalid UE golomb code
    Last message repeated 29 times
Invalid UE golomb code
    Last message repeated 29 times
Invalid UE golomb code
    Last message repeated 29 times
Invalid UE golomb code
    Last message repeated 29 times
Invalid UE golomb code
    Last message repeated 29 times
Invalid UE golomb code
    Last message repeated 29 times
Invalid UE golomb code


  


  Why does ffmpeg cares if its stdout is a file or a pipe ?

  


• rtmpproto : Validate the embedded flv packet size before copying

  3 octobre 2013, par Martin Storsjö

  rtmpproto : Validate the embedded flv packet size before copying
  This wasn’t an issue prior to 58404738, when the whole RTMP packet
  
  was copied at once and the length of the individual embedded flv
  
  packets only were validated by the flv demuxer.
  Prior to this patch, this could lead to reads and writes out of bound.
  Signed-off-by : Martin Storsjö <martin@martin.st>
  

  • [DH] libavformat/rtmpproto.c

• How to detect a common scene in a set of videos with ffmpeg

  6 août 2019, par Hans J

  I have a set of videos that are assumed to contain a common (or very similar) scene. I want to be able to detect (with FFmpeg) what that common scene is, how long the scene is, and where the scene is in each individual video.

  The scene would be assumed to be longer than 10 seconds (This is an arbitrary choice, it can be changed).

  The final output of the command would include the various time-codes of the instance of the scene in each video. Assuming a timebase 1/1, with a common scene that is 60 seconds long, an output would along the lines of :

  Video1.mp4 0 60
  
  Video2.mp4 120 180
  
  Video3.mp4 50 110
  
  Video4.mp4 null

  where video4 does not contain any common scene.

  For example, I could have three episodes of a TV show. They all contain the same commercial. Without knowing what that commercial is, I want to be able to find where that commercial shows up in each of the episodes.

  Note : For the purpose of a good solution, the common scenes do not have to exactly match. Because there could be artifacts or embedded subtitles in one episode and not the other.