Recherche avancée

Sur d’autres sites (6336)

• lavfi : add filter dnn_detect for object detection

  7 février 2021, par Guo, Yejun

  lavfi : add filter dnn_detect for object detection
  Below are the example steps to do object detection :
  1. download and install l_openvino_toolkit_p_2021.1.110.tgz from
  
  https://software.intel.com/content/www/us/en/develop/tools/openvino-toolkit/download.html
  
    or, we can get source code (tag 2021.1), build and install.
  
  2. export LD_LIBRARY_PATH with openvino settings, for example :
  
  .../deployment_tools/inference_engine/lib/intel64/ :.../deployment_tools/inference_engine/external/tbb/lib/
  
  3. rebuild ffmpeg from source code with configure option :
  
  — enable-libopenvino
  
  — extra-cflags='-I.../deployment_tools/inference_engine/include/'
  
  — extra-ldflags='-L.../deployment_tools/inference_engine/lib/intel64'
  
  4. download model files and test image
  
  wget https://github.com/guoyejun/ffmpeg_dnn/raw/main/models/openvino/2021.1/face-detection-adas-0001.bin
  
  wget https://github.com/guoyejun/ffmpeg_dnn/raw/main/models/openvino/2021.1/face-detection-adas-0001.xml
  
  wget
  
  https://github.com/guoyejun/ffmpeg_dnn/raw/main/models/openvino/2021.1/face-detection-adas-0001.label
  
  wget https://github.com/guoyejun/ffmpeg_dnn/raw/main/images/cici.jpg
  
  5. run ffmpeg with :
  
  ./ffmpeg -i cici.jpg -vf dnn_detect=dnn_backend=openvino:model=face-detection-adas-0001.xml:input=data:output=detection_out:confidence=0.6:labels=face-detection-adas-0001.label,showinfo -f null -
  We'll see the detect result as below :
  
  [Parsed_showinfo_1 @ 0x560c21ecbe40]   side data - detection bounding boxes :
  
  [Parsed_showinfo_1 @ 0x560c21ecbe40] source : face-detection-adas-0001.xml
  
  [Parsed_showinfo_1 @ 0x560c21ecbe40] index : 0,  region : (1005, 813) -> (1086, 905), label : face, confidence : 10000/10000.
  
  [Parsed_showinfo_1 @ 0x560c21ecbe40] index : 1,  region : (888, 839) -> (967, 926), label : face, confidence : 6917/10000.
  There are two faces detected with confidence 100% and 69.17%.
  Signed-off-by : Guo, Yejun <yejun.guo@intel.com>
  

  • [DH] configure
  • [DH] doc/filters.texi
  • [DH] libavfilter/Makefile
  • [DH] libavfilter/allfilters.c
  • [DH] libavfilter/vf_dnn_detect.c

• How to fix a webm file without audio ?

  14 juillet 2022, par John

  I use mediarecorder to record video and audio from a user's browser. We record every 15 seconds and then upload that blog to S3. Then we combine all the files together to make one webm file. I believe the first file isn't right because when I combine the files, there is not any audio - only video.

  


  Is there a way to alter the headers in the first file to use the audio in all of the subsequent files ? OR is there an FFMPEG command to force using the audio ? I know they exist in the other files.

  


  I don't believe this is important but here is the code that I use to save and combine the webm blobs.

  


  First I save the blobs from the media recorder

  


    recorder = new MediaRecorder(local_media_stream.remoteStream, {
               mimeType: encoding_options,
               audioBitsPerSecond: 96000,
               videoBitsPerSecond: bits_per_second,
            });
  recorder.ondataavailable = function(e) {
          that.save_blob(e.data, blob_index);      
         }


  


  Then later I combine each of those blobs.

  


  bucket = Aws::S3::Resource.new(region:'us-east-1').bucket("files")

keys = bucket.objects(prefix: "files").collect(&:key)

temp_webm_file = Tempfile.new(['total', '.webm'])
keys.each_with_index do |key, index|
    temp_webm_file.write bucket.object(key).get.body.read
end
temp_webm_file.close()


  


  One thing I know that fixes the issue is if I combine a short webm file with audio to the very beginning. Then the audio all works.

  


• Missing piece between libjpeg-turbo & h264 ffmpeg C/C++

  15 octobre 2022, par Nelstaar

  On the left side I have a buffer with decoded pixels that I can get in two formats :

  


  RGB interleaved/packed where bytes in buffer are R0G0B0R1G1B1....

  


  or

  


  YUV444 interleaved/packed where bytes in buffer are Y0U0V0Y1U1V1...

  


  (JCS_RGB or JCS_YCbCr in jpeglib.h)

  


  (Please note that I use libjpeg-turbo because I need to decompress a cropped region of the image. (jpeg_crop_scanline()))

  


  On the right side I have x264 codec via ffmpeg that support only planar pixel formats :

  


  yuv420p, yuvj420p, yuv422p, yuvj422p, yuv444p, yuvj444p, nv12, nv16, nv21, yuv420p10le, yuv422p10le, yuv444p10le, nv20le

  


  yuv444p where bytes in buffer are Y0Y1Y2...U0U1...V0V1...

  


  according to ffmpeg -h encoder=libx264

  


  I have some ideas already :

  


  

  • Decompress Jpeg to RBG888 in buffer 1 then libswscale to yuv420p in buffer 2 and encoding. (copy)

  


  • Decompress Jpeg to YUV444 interleaved in buffer 1 then SSSE3 magic in buffer 1 to yuv444p and encoding. (no copy)

  


  • or else.

  


  


  What would be the most effective fastest way ?

  


  I which to avoid buffer copy.

  


  Movie have the same width & height than Jpegs.