Recherche avancée

Sur d’autres sites (13947)

• ffmpeg compilation problem : avcodec_find_decoder always returns null

  15 mars 2016, par Adion

  I recently tried to upgrade the ffmpeg libraries I use in my Mac OS X application by downloading and compiling ffmpeg from source.

  My code works correctly with pre-compiled libraries of the same version on windows.
  On Mac OS X, the library appears to work (it can open the file and find the streams and codecs used), but when it gets to avcodec_find_decoder, this function always returns null.

  The code has worked with an older version of the library (compiled a year ago on Mac OS X 10.5)

  I configured fmpeg using

  ./configure --extra-cflags="-arch i386" --extra-ldflags='-arch i386' --arch=x86_32 --target-os=darwin --enable-cross-compile --disable-indev=jack --enable-shared --disable-static

  I checked config.mak, and it appears to have the decoders for the file types I tried enabled (ogg, vorbis, avi, mkv, ...)
  I also checked that the correct header files have been used and that the newly compiled library is used.

  I have found only some older posts relating to this issue, but without any solution :

  http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2007-January/021399.html

  http://libav-users.943685.n4.nabble.com/avcodec-find-decoder-problem-td944800.html

  Edit : checking further, it appears av_codec_next(NULL) returns null as well, which means there isn’t a single codec available, or that first_avcodec in utils.c is not set (I actually haven’t found at all where this variable is set, I would have assumed av_register_all, but I can’t find it there)

• aarch64 : vp9 : use alternative returns in the core loop filter function

  14 novembre 2016, par Janne Grunau

  aarch64 : vp9 : use alternative returns in the core loop filter function
  Since aarch64 has enough free general purpose registers use them to
  
  branch to the appropiate storage code. 1-2 cycles faster for the
  
  functions using loop_filter 8/16, ... on a cortex-a53. Mixed results
  
  (up to 2 cycles faster/slower) on a cortex-a57.
  

  • [DBH] libavcodec/aarch64/vp9lpf_neon.S

• tf.contrib.signal.stft returns an empty matrix

  9 décembre 2017, par matt-pielat

  This is the piece of code I run :

  import tensorflow as tf
  
  
  
  sess = tf.InteractiveSession()
  
  
  
  filename = 'song.mp3' # 30 second mp3 file
  
  SAMPLES_PER_SEC = 44100
  
  
  
  audio_binary = tf.read_file(filename)
  
  
  
  pcm = tf.contrib.ffmpeg.decode_audio(audio_binary, file_format='mp3', samples_per_second=SAMPLES_PER_SEC, channel_count = 1)
  
  stft = tf.contrib.signal.stft(pcm, frame_length=1024, frame_step=512, fft_length=1024)
  
  
  
  sess.close()

  The mp3 file is properly decoded because print(pcm.eval().shape) returns :

  (1323119, 1)

  And there are even some actual non-zero values when I print them with print(pcm.eval()[1000:1010]) :

  [[ 0.18793298]
  
   [ 0.16214484]
  
   [ 0.16022217]
  
   [ 0.15918455]
  
   [ 0.16428113]
  
   [ 0.19858395]
  
   [ 0.22861415]
  
   [ 0.2347789 ]
  
   [ 0.22684409]
  
   [ 0.20728172]]

  But for some reason print(stft.eval().shape) evaluates to :

  (1323119, 0, 513) # why the zero dimension?

  And therefore print(stft.eval()) is :

  []

  According to this the second dimension of the tf.contrib.signal.stft output is equal to the number of frames. Why are there no frames though ?