Recherche avancée

Sur d’autres sites (9318)

• Revision 64742f825d : Merge "Elevate NEWMV mode checking threshold in real time"

  2 juillet 2014, par Yunqing Wang

  Merge "Elevate NEWMV mode checking threshold in real time"

• Capturing and saving an entire RTP video stream using FFmpeg

  27 juin 2021, par Alexander

  Question

  


  How to capture an entire RTP video stream and save it to a file - without defining the duration, shorter than the original video sequence ?

  


  Background

  


  I am streaming a video sequence over the network with FFmpeg and RTP (football (b) at Xiph) using the following command :

  


  ffmpeg -re -i football.y4m -preset slow -c:v libx264 -bf 0 -g 10 -x264opts scenecut=0  -f rtp -sdp_file video.sdp "rtp://<ip>:<port>"</port></ip>

  


  Assuming the destination machine has got access to its own copy of video.sdp, I then capture and save the stream using :

  


  ffmpeg -protocol_whitelist file,rtp,udp -i video.sdp -t 00:00:08 <out>.mp4</out>

  


  My concern is that the time -t 00:00:08 must always be defined in order to terminate capturing. Moreover, if it is not shorter than the original video sequence (8.67 s), capturing is not automatically terminated.

  


  I do not encounter the issue when visualising the received stream with (the entire video is played) :

  


  ffplay -protocol_whitelist file,rtp,udp -i video.sdp

  


• Broken Pipe Error [Errno 32] after saving more than ten PNG images using PIL

  6 septembre 2019, par Blinko

  I am using the ffmpeg pipeline to compress a series of png frames and output an mp4 video, while also saving each frame as a separate png image. The problem is that every time I run the script, I get a Broken Pipe error on the 11th image :

  Traceback (most recent call last):
  
    File "/anaconda3/envs/tensorflow/lib/python3.6/site-packages/PIL/ImageFile.py", line 487, in _save
  
      fh = fp.fileno()
  
  AttributeError: '_idat' object has no attribute 'fileno'
  
  
  
  During handling of the above exception, another exception occurred:
  
  
  
  Traceback (most recent call last):
  
    File "gen_clip.py", line 173, in <module>
  
      main()
  
    File "gen_clip.py", line 167, in main
  
      ts_pos = (0,frame.samplesperbeam-50)
  
    File "/Users/Amanda/Desktop/data/pyARIS.py", line 1028, in make_video
  
      im.save(pipe.stdin, 'PNG')
  
    File "/anaconda3/envs/tensorflow/lib/python3.6/site-packages/PIL/Image.py", line 1994, in save
  
      save_handler(self, fp, filename)
  
    File "/anaconda3/envs/tensorflow/lib/python3.6/site-packages/PIL/PngImagePlugin.py", line 868, in _save
  
      [("zip", (0, 0)+im.size, 0, rawmode)])
  
    File "/anaconda3/envs/tensorflow/lib/python3.6/site-packages/PIL/ImageFile.py", line 502, in _save
  
      fp.write(d)
  
    File "/anaconda3/envs/tensorflow/lib/python3.6/site-packages/PIL/PngImagePlugin.py", line 730, in write
  
      self.chunk(self.fp, b"IDAT", data)
  
    File "/anaconda3/envs/tensorflow/lib/python3.6/site-packages/PIL/PngImagePlugin.py", line 717, in putchunk
  
      fp.write(data)
  
  BrokenPipeError: [Errno 32] Broken pipe
  
  </module>

  I am not too familiar with PIL or ffmpeg, so I am not sure whether it’s a problem with the pipe parameters that I am passing in or something to do with the code itself. No matter what the starting frame number is, the error always appears after 10 images are saved.

  Here is the array of pipe parameters I’m sending to the command prompt :

  command = ['ffmpeg',
  
         '-y',
  
         '-f', 'image2pipe',
  
         '-vcodec', 'png',
  
         '-r', '1',
  
         '-r', str(fps), # frames per second
  
         '-i', '-', # The input comes from a pipe
  
         '-an', # Tells FFMPEG not to expect any audio
  
         '-vcodec', 'mpeg4',
  
         '-b:v', '5000k',
  
         '../'+filename+"/"+filename+".mp4",
  
         '-hide_banner',
  
         '-loglevel', 'panic']

  This is the loop that generates and saves each image :

      for i in tqdm.tqdm(range(start_frame, end_frame)):
  
          frame = FrameRead(data, i)
  
          frame_image = np.zeros([ydim, xdim], dtype=np.uint8)
  
          frame_image[image_write_rows, image_write_cols] = frame.frame_data[sample_read_rows, sample_read_cols]
  
  
  
          all_frame_data.append(frame.frame_data)
  
          im = Image.fromarray(cm(frame_image, bytes=True))
  
  
  
          if timestamp == True:
  
              ts = str(datetime.datetime.fromtimestamp(frame.sonartimestamp/1000000, pytz.timezone('UTC')).strftime('%Y-%m-%d %H:%M:%S'))
  
              text = "%s\n%d" % (ts, i)
  
              draw = ImageDraw.Draw(im)
  
              draw.text(ts_pos,text,font=font, fill = 'white')
  
  
  
          im.save(directory+filename+'/frames/'+str(i)+'_'+str(frame.sonartimestamp)+'_'+filename+'.png'  , "PNG")
  
          im.save(pipe.stdin, 'PNG')
  
          j += 1

  I am running this on MacOS, Python 3.6. Help on the cause of the error and/or possible solutions will be greatly appreciated !