Recherche avancée

Sur d’autres sites (8322)

• tools/python : add script to convert TensorFlow model (.pb) to native model (.model)

  13 juin 2019, par Guo, Yejun

  tools/python : add script to convert TensorFlow model (.pb) to native model (.model)
  For example, given TensorFlow model file espcn.pb,
  
  to generate native model file espcn.model, just run :
  
  python convert.py espcn.pb
  In current implementation, the native model file is generated for
  
  specific dnn network with hard-code python scripts maintained out of ffmpeg.
  
  For example, srcnn network used by vf_sr is generated with
  
  https://github.com/HighVoltageRocknRoll/sr/blob/master/generate_header_and_model.py#L85
  In this patch, the script is designed as a general solution which
  
  converts general TensorFlow model .pb file into .model file. The script
  
  now has some tricky to be compatible with current implemention, will
  
  be refined step by step.
  The script is also added into ffmpeg source tree. It is expected there
  
  will be many more patches and community needs the ownership of it.
  Another technical direction is to do the conversion in c/c++ code within
  
  ffmpeg source tree. While .pb file is organized with protocol buffers,
  
  it is not easy to do such work with tiny c/c++ code, see more discussion
  
  at http://ffmpeg.org/pipermail/ffmpeg-devel/2019-May/244496.html. So,
  
  choose the python script.
  Signed-off-by : Guo, Yejun <yejun.guo@intel.com>
  

  • [DH] .gitignore
  • [DH] tools/python/convert.py
  • [DH] tools/python/convert_from_tensorflow.py

• Making a ffmpeg screen capture on Mac OS X using YUV 4:2:0 Planar color model

  30 mai 2019, par Bass

  I make screen recordings with ffmpeg, using avfoundation on Mac OS X, x11grab on Linux and gdigrab on Windows.

  The resulting files should be compatible with modern web browsers (<video></video>), so I use H.264 codec and request YUV 4:2:0 Planar pixel format.

  On Mac OS X, however (unlike Linux and Windows), I receive the following logging :

  /usr/local/bin/ffmpeg -y -v error -f avfoundation -threads 0 -hide_banner -i 1:none -f mp4 -vcodec h264 -pix_fmt yuv420p -r 25/1 -qscale:v 1 -vf scale=-1:1080 target.mp4
  
  [avfoundation @ 0x7fdba2003a00] Selected pixel format (yuv420p) is not supported by the input device.
  
  [avfoundation @ 0x7fdba2003a00] Supported pixel formats:
  
  [avfoundation @ 0x7fdba2003a00]   uyvy422
  
  [avfoundation @ 0x7fdba2003a00]   yuyv422
  
  [avfoundation @ 0x7fdba2003a00]   nv12
  
  [avfoundation @ 0x7fdba2003a00]   0rgb
  
  [avfoundation @ 0x7fdba2003a00]   bgr0

  Still, according to mplayer, the resulting MP4 file seems to have YUV 4:2:0 Planar color model :

  [h264 @ 0x1048a8ac0]Format yuv420p chosen by get_format().
  
  [h264 @ 0x1048a8ac0]Reinit context to 1728x1088, pix_fmt: yuv420p
  
  [h264 @ 0x1048a8ac0]Format yuv420p chosen by get_format().
  
  [h264 @ 0x1048a8ac0]Reinit context to 1728x1088, pix_fmt: yuv420p
  
  [swscaler @ 0x1048c3cc0]bicubic scaler, from yuv420p to yuyv422 using MMXEXT
  
  *** [scale] Exporting mp_image_t, 1728x1080x12bpp YUV planar, 2799360 bytes
  
  *** [vo] Allocating mp_image_t, 1728x1080x16bpp YUV packed, 3732480 bytes

  the same confirmed by ffmpeg :

  $ ffmpeg -i target.mp4 -hide_banner
  
  Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'target.mp4':
  
    Metadata:
  
      major_brand     : isom
  
      minor_version   : 512
  
      compatible_brands: isomiso2avc1mp41
  
      encoder         : Lavf58.20.100
  
    Duration: 00:00:04.72, start: 0.000000, bitrate: 201 kb/s
  
      Stream #0:0(und): Video: h264 (High) (avc1 / 0x31637661), yuv420p, 1728x1080, 197 kb/s, 25 fps, 25 tbr, 12800 tbn, 50 tbc (default)
  
      Metadata:
  
        handler_name    : VideoHandler

  Questions :

  1. Can someone explain the above ffmpeg logging ?
  2. If I still need to convert the avfoundation video stream to yuv420p, how do I make it on the fly (in a single ffmpeg pass) ?

• libavfilter/dnn : remove limit for the name of DNN model input/output

  25 avril 2019, par Guo, Yejun

  libavfilter/dnn : remove limit for the name of DNN model input/output
  remove the requirment that the name of DNN model input/output
  
  should be "x"/"y",
  Signed-off-by : Guo, Yejun <yejun.guo@intel.com>
  
  Signed-off-by : Pedro Arthur <bygrandao@gmail.com>
  

  • [DH] libavfilter/dnn_backend_native.c
  • [DH] libavfilter/dnn_backend_tf.c
  • [DH] libavfilter/dnn_interface.h
  • [DH] libavfilter/vf_sr.c