Recherche avancée

Sur d’autres sites (5641)

• Segmentation Fault error when trying to compare two videos with pHash library and its ruby bindings

  10 juillet 2014, par anado1771

  I have set up my system with the latest ffmpeg and pHash libraries (ffmpeg-2.2.1 and pHash-0.9.6) as well as the pHash ruby gem (https://github.com/toy/pHash).

  I am using ruby and attempting to compare two video files like this :

  require 'phash/video'
  
  
  
  video1 = Phash::Video.new('video1.mp4')
  
  video2 = Phash::Video.new('video2.mp4')
  
  
  
  video1 % video2

  Executing this script results in a Segmentation fault :

  ..../gems/pHash-1.1.4/lib/phash/video.rb:20: [BUG] Segmentation fault
  
  ruby 1.9.3p545 (2014-02-24 revision 45159) [x86_64-darwin13.1.0]
  
  
  
  -- Control frame information -----------------------------------------------
  
  c:0008 p:---- s:0029 b:0029 l:000028 d:000028 CFUNC  :ph_dct_videohash
  
  c:0007 p:0042 s:0024 b:0024 l:000023 d:000023 METHOD .../gems/pHash-1.1.4/lib/phash/video.rb:20
  
  c:0006 p:0038 s:0017 b:0017 l:000016 d:000016 METHOD .../gems/pHash-1.1.4/lib/phash.rb:43
  
  c:0005 p:0025 s:0014 b:0014 l:000013 d:000013 METHOD .../gems/pHash-1.1.4/lib/phash.rb:39
  
  c:0004 p:0011 s:0011 b:0011 l:000010 d:000010 METHOD .../gems/pHash-1.1.4/lib/phash.rb:48
  
  c:0003 p:0050 s:0006 b:0006 l:000128 d:0011b8 EVAL   video_test_phash.rb:3
  
  c:0002 p:---- s:0004 b:0004 l:000003 d:000003 FINISH
  
  c:0001 p:0000 s:0002 b:0002 l:000128 d:000128 TOP   
  
  
  
  -- Ruby level backtrace information ----------------------------------------
  
  video_test_phash.rb:3:in `<main>'
  
  .../gems/pHash-1.1.4/lib/phash.rb:48:in `similarity'
  
  .../gems/pHash-1.1.4/lib/phash.rb:39:in `phash'
  
  .../gems/pHash-1.1.4/lib/phash.rb:43:in `compute_phash'
  
  .../gems/pHash-1.1.4/lib/phash/video.rb:20:in `video_hash'
  
  .../gems/pHash-1.1.4/lib/phash/video.rb:20:in `ph_dct_videohash'
  
  
  
  ...
  
  
  
  Abort trap: 6
  
  </main>

  It appears that the crash happens in the ph_dct_videohash function which is part of the pHash library. The function is in file pHash.cpp. I am copying it here in case it would make sense to someone :

  ulong64* ph_dct_videohash(const char *filename, int &Length){
  
  
  
      CImgList *keyframes = ph_getKeyFramesFromVideo(filename);
  
      if (keyframes == NULL)
  
      return NULL;
  
  
  
      Length = keyframes->size();
  
  
  
      ulong64 *hash = (ulong64*)malloc(sizeof(ulong64)*Length);
  
      CImg<float> *C = ph_dct_matrix(32);
  
      CImg<float> Ctransp = C->get_transpose();
  
      CImg<float> dctImage;
  
      CImg<float> subsec;
  
      CImg currentframe;
  
  
  
      for (unsigned int i=0;i &lt; keyframes->size(); i++){
  
      currentframe = keyframes->at(i);
  
      currentframe.blur(1.0);
  
      dctImage = (*C)*(currentframe)*Ctransp;
  
      subsec = dctImage.crop(1,1,8,8).unroll('x');
  
      float med = subsec.median();
  
      hash[i] =     0x0000000000000000;
  
      ulong64 one = 0x0000000000000001;
  
      for (int j=0;j&lt;64;j++){
  
          if (subsec(j) > med)
  
          hash[i] |= one;
  
          one = one &lt;&lt; 1;
  
      }
  
      }
  
  
  
      keyframes->clear();
  
      delete keyframes;
  
      keyframes = NULL;
  
      delete C;
  
      C = NULL;
  
      return hash;
  
  }
  
  </float></float></float></float>

  Any help is very much appreciated !

• ffmpeg static build produces huge files

  3 mai 2014, par user3581244

  I’ve recently managed to get a static build of ffmpeg installed on an EC2 instance :

  http://ffmpeg.gusari.org/static/64bit/ffmpeg.static.64bit.2014-03-05.tar.gz

  But when I perform this command on a video file :

  ffmpeg -ss 00:00:00 -i VIDEO0031.mp4 -r 8 -vframes 32 -vbsf remove_extra -an -vcodec pam -f rawvideo -y test_002.pam

  It produces a file that is 190MB which is huge and then it messes up the other commands I use to convert it into a gif file.

  Does anyone know if this is just an issue with using a static build ? I didn’t experience this issue using an older ffmpeg install ( v0.6 I believe ) so am tempted to go back to the old install but I also need to have ffprobe installed so am worried this would cause issues.

  A bit of a random question but I thought it was worth asking just in case someone had come across a similar thing.

  Thanks !

• Revision 81ad047ee5 : VP8 for ARMv8 by using NEON intrinsics 06 Add idct_dequant_full_2x_neon.c - idc

  17 décembre 2013, par James Yu

  Changed Paths :
   Delete /vp8/common/arm/neon/idct_dequant_full_2x_neon.asm

  
   Add /vp8/common/arm/neon/idct_dequant_full_2x_neon.c

  
   Modify /vp8/vp8_common.mk

  
  
  VP8 for ARMv8 by using NEON intrinsics 06

  Add idct_dequant_full_2x_neon.c
   idct_dequant_full_2x_neon

  ==== Summary of apply VP8 decode patch series ====
  Benchmark on Samsung Chromebook, Cortex-A15, 1.7GHz, Dual core
  Toolchain : linaro-1.13.1-4.8-2014.01
  Compile argument : CROSS=arm-linux-gnueabihf- ../libvpx/configure
  —target=armv7-linux-gcc —prefix=$HOME/out
  —enable-shared —cpu=cortex-a7
  Test argument : vpxdec —summary —noblit ./tears_of_steel_1080p.webm

  NEON assembly 46.68 (fps)
  Apply patch 06 46.65, -0.03
  Apply patch 07 46.86, +0.21
  Apply patch 08 46.58, -0.28
  Apply patch 09 46.57, -0.01
  Apply patch 10 46.51, -0.06
  Apply patch 11 46.13, -0.38
  Apply patch 12 45.42, -0.71
  Apply patch 13 46.06, +0.64
  Apply patch 14 45.19, -0.87
  Apply patch 15 45.93, +0.74
  Apply patch 16 45.48, -0.45
  Apply patch 17 45.84, +0.36
  Apply patch 18 45.91, +0.07 <= With all NEON intrinsics patches
  Total -0.77 fps, 1.65% performance regression

  Change-Id : I77bfc9eaccfb97b8d401e949ceff8795e26ca6b7
  Signed-off-by : James Yu <james.yu@linaro.org>