Recherche avancée

Sur d’autres sites (7628)

• How can I create a c++ console application which makes use of an open source c project

  2 octobre 2018, par Bogdan Daniel

  I’ve been playing around with a compiled version of https://github.com/FFmpeg/FFmpeg. But it has some problems, during the closing of an opened stream using avio_close(it takes a really long time to close it).

  I’ve been trying to understand what could go wrong by reading through the implementation, but couldn’t find anything.

  What I would like to do, is to actually have a C++ console application which uses the c files and to debug them while running the code(using breakpoints and so on).

  Unfortunately I cannot find any information on how to set it up. Simply copying the files in a new console application is not enough.

  Thanks in advance for any suggestions.

  Edit : I can already see a closing vote. If this question is not appropriate for this website, I will delete it. But please point me in the right direction of where to post it.Although it is quite a general question, I believe that it is clear and enough information has been provided.

  Edit2 : Yes, I was a bit unclear about what I’m using.

  I’m using Visual Studio Community 2017 on Windows 10 for creating the C++ console application.

  Edit3 :

  Steps that I’ve taken into using the source files into my console application which was using the DLLs.

  1. Copy all of the contents of the FFmpeg-master to my console application
  2. Include all of them in the c++ console application project
  3. Run a build - takes forever and has infinite build errors( > 1000) - probably compiler related

  Edit4 :

  I have no actual errors with avio_close, it just takes too long to close the stream. What I found out when playing around with the settings is that when the fifo_size parameter is set, closing the stream is a lot faster depending on how small the set value is. And it sort of makes sense since fifo_size is related to the packet size, but I haven’t found out where in the code this size has an impact.

  fifo_size=units
  Set the UDP receiving circular buffer size, expressed as a number of packets with size of 188 bytes. If not specified defaults to 7*4096.

  Edit5 : I still haven’t found a way to compile the open source ffmpeg project into libs, dlls and pdbs. Am I the first one needing such files(seems unrealistic) ?
  The only tool capable of delivering those so far is vcpkg. The only problem is that it compiles the 3.3.3 version and the latest is 4.0.2 .

  I tried to modify the vcpkg\ports\ffmpeg\portfile.cmake file to include the latest version of ffmpeg, but it doesn’t build it.

  Are there any other suggestions ?

• Ubuntu 16.04 + OpenCV + FFMPEG + CUDA don't open default camera

  20 juillet 2018, par Newton Pasqualini Filho

  I am trying to work with accelerated OpenCV using my own build of FFMPEG with CUDA support for GPU video processing using Python.

  My application must support any kind of video devices, including USB and IP cameras.

  I have a config file parsed with ConfigParser that loads the camera device path, as described in OpenCV documentation we can access usb camera by index number and for IP camera we use RTSP protocol so the path could be sometimes 0 or rtsp://192.168.1.2/0.

  To force the use of FFMPEG behind OpenCV I am trying to open the video capture stream like bellow :

  video_capture = cv.VideoCapture()
  
  stream_opened = video_capture.open(camera_path, cv.CAP_FFMPEG)

  The problem is when I try to open the default USB camera camera_path = 0 the video_capture.open method always returns false, but when it is a RTSP like camera_path = "rtsp://192.168.1.2/1" it open the video capture stream ok.

  When I let OpenCV decide what backend will be used it calls to GStreammer and it opens the capture stream like this :

  video_capture = cv.VideoCapture(camera_path)

  I suspect that FFMPEG does not have any support for capture video from a USB camera, does it ? Is there some missing feature to my ffmpeg build ?

  ffmpeg version N-91487-g1809f1c Copyright (c) 2000-2018 the FFmpeg developers
  
  built with gcc 5.4.0 (Ubuntu 5.4.0-6ubuntu1~16.04.10) 20160609
  
  configuration: --prefix=/usr --pkg-config-flags=--static --enable-cuda-sdk --enable-cuvid --enable-libnpp --extra-cflags=-I/usr/local/cuda/include/ --extra-ldflags=-L/usr/local/cuda/lib64/ --nvccflags='-gencode arch=compute_61,code=sm_61 -O2' --enable-gpl --enable-libass --enable-libfdk-aac --enable-libx264 --extra-libs=-lpthread --enable-libx265 --enable-nvenc --enable-nonfree
  
  libavutil      56. 18.102 / 56. 18.102
  
  libavcodec     58. 21.105 / 58. 21.105
  
  libavformat    58. 17.101 / 58. 17.101
  
  libavdevice    58.  4.101 / 58.  4.101
  
  libavfilter     7. 26.100 /  7. 26.100
  
  libswscale      5.  2.100 /  5.  2.100
  
  libswresample   3.  2.100 /  3.  2.100
  
  libpostproc    55.  2.100 / 55.  2.100
  
  Hyper fast Audio and Video encoder
  
  usage: ffmpeg [options] [[infile options] -i infile]... {[outfile options] outfile}...

• Opencv 3.1 compiled with FFMPEG, but won't open https urls

  2 mars 2018, par stalagmite7

  I work on a 16.04 system, and have successfully installed opencv 3.1 with FFMPEG flags enabled. I double checked this was actually the case by cv2.getBuildInformation() and I got FFMPEG = YES.

  I am trying to open a video that is hostel on a private server by my workplace (I am logged in to the VPN, in case thats a concern) and I can access this video over the browser. But videocapture with cv2 fails.

  >>> cap = cv2.VideoCapture("https://xxx.mp4", cv2.CAP_ANY) #dummy url
  
  >>> cap
  
  <videocapture 0x7f63300fa4b0="0x7f63300fa4b0">
  
  >>> cap.isOpened()
  
  False
  
  </videocapture>

  This is always the case for https urls. It seems to be able to work with local videos just fine. I have tried a bunch of different thing : initially thought it was a gstreamer problem so I checked my plugins, had some gst-bad versions (ref : https://github.com/GStreamer/gst-plugins-ugly), removed those and replaced with good versions, no joy. Also tried to explicitly tell videoCapture to use cv2.CAP_ANY and cv2.CAP_FFMPEG flags while reading the video, still no luck.

  I disabled the Gstreamer flag while compiling opencv, but even with it set to ON, there was no difference in my problem.

  I haven’t been able to find a solution to this issue and have been looking and trying different things for days now ! Any ideas ?