Recherche avancée

Sur d’autres sites (9654)

• ffmpeg - seamless crossfade loop for the part of video

  14 janvier 2021, par Flamin GO

  I need to apply crossfade to the last X frames of a video with the first X frames in order to obtain a seamless loop, but making this for the necessary part of video.

  


  Here's the answer for looping the entire video.

  


  Currently what I have :
(Whole video duration = 25. Cutted (result) part = 15 sec (from 5 to 20 sec pos). Transition = 1 sec.)

  


  ffmpeg -i input.mp4 -ss 5 -to 20 -filter_complex
    "[0]split[body][pre];
     [pre]trim=duration=1,format=yuva420p,fade=d=1:alpha=1,setpts=PTS+( (15+(5-1)) /TB)[jt];
     [body]trim=1,setpts=PTS-STARTPTS[main];
     [main][jt]overlay"  -c:v libx264 -preset veryslow -b:v 2500K output.mp4
 


  


  In this case, everything works, but at the end of the resulting video, a piece from the original video is superimposed, which starts from 0 to 1 second, and not from 4 to 5 seconds of the original video, as it should be.

  


  I read the official ffmpeg documentation, tried some actions on "start/end" parameters for "trim/fade" with changing of "setpts", but I always got just another batch of bugs.

  


• Why the output of the ffmpeg-python doesn't match the image shape ?

  9 novembre 2019, par Swi Jason

  I used the ffmpeg-python module to convert video to images. Specifically, I used the code provided by the official git repo of ffmpeg-python, as below

  out, _ = (
  
      ffmpeg
  
      .input(in_filename)
  
      .filter('select', 'gte(n,{})'.format(frame_num))
  
      .output('pipe:', vframes=1, format='image2', vcodec='mjpeg')
  
      .run(capture_stdout=True)
  
  )
  
  im = np.frombuffer(out, 'uint8')
  
  print(im.shape[0]/3/1080)
  
  # 924.907098765432

  The original video is of size (1920, 1080) and pix_fmt ’yuv420p’, but the outputs of the above code is not 1920.

  I have figured out by myself that the output of ffmpeg.run() is not a decoded image array, but a byte string encoded by JPEG format. To restore the image into a numpy array, simply use the cv2.imdecode() function. For example,

  im = cv2.imdecode(im, cv2.IMREAD_COLOR)

  However, I can’t use opencv on my embeded Linux system. So my question now is that, can I get numpy output from ffmpeg-python directly, without the need of converting it by opencv ?

• libavfi/dnn : add LibTorch as one of DNN backend

  15 mars 2024, par Wenbin Chen

  libavfi/dnn : add LibTorch as one of DNN backend
  PyTorch is an open source machine learning framework that accelerates
  
  the path from research prototyping to production deployment. Official
  
  website : https://pytorch.org/. We call the C++ library of PyTorch as
  
  LibTorch, the same below.
  To build FFmpeg with LibTorch, please take following steps as
  
  reference :
  
  1. download LibTorch C++ library in
  
   https://pytorch.org/get-started/locally/,
  
  please select C++/Java for language, and other options as your need.
  
  Please download cxx11 ABI version :
  
   (libtorch-cxx11-abi-shared-with-deps-*.zip).
  
  2. unzip the file to your own dir, with command
  
  unzip libtorch-shared-with-deps-latest.zip -d your_dir
  
  3. export libtorch_root/libtorch/include and
  
  libtorch_root/libtorch/include/torch/csrc/api/include to $PATH
  
  export libtorch_root/libtorch/lib/ to $LD_LIBRARY_PATH
  
  4. config FFmpeg with ../configure —enable-libtorch \
  
   —extra-cflag=-I/libtorch_root/libtorch/include \
  
   —extra-cflag=-I/libtorch_root/libtorch/include/torch/csrc/api/include \
  
   —extra-ldflags=-L/libtorch_root/libtorch/lib/
  
  5. make
  To run FFmpeg DNN inference with LibTorch backend :
  
  ./ffmpeg -i input.jpg -vf \
  
  dnn_processing=dnn_backend=torch:model=LibTorch_model.pt -y output.jpg
  The LibTorch_model.pt can be generated by Python with torch.jit.script()
  
  api. https://pytorch.org/tutorials/advanced/cpp_export.html. This is
  
  pytorch official guide about how to convert and load torchscript model.
  
  Please note, torch.jit.trace() is not recommanded, since it does
  
  not support ambiguous input size.
  Signed-off-by : Ting Fu <ting.fu@intel.com>
  
  Signed-off-by : Wenbin Chen <wenbin.chen@intel.com>
  
  Reviewed-by : Guo Yejun <yejun.guo@intel.com>
  

  • [DH] configure
  • [DH] libavfilter/dnn/Makefile
  • [DH] libavfilter/dnn/dnn_backend_torch.cpp
  • [DH] libavfilter/dnn/dnn_interface.c
  • [DH] libavfilter/dnn_filter_common.c
  • [DH] libavfilter/dnn_interface.h
  • [DH] libavfilter/vf_dnn_processing.c