Optimized OpenCV 3.4.2 + Python3 (with GUI and Video)

Here's a container image based on debian:jessie with:

• OpenCV 3.4.2
• Python 3.4
• GUI libraries (e.g. for cv2.imshow())
• Optimizations (Lapack, Eigen, IPP)
• Video grabbing libraries (ffmpeg, v4l, GStreamer)

It enables you to run any code that uses the OpenCV library, even GUI code, on any OS that can run Docker - therefore you now can run your OpenCV programs on almost any OS, without spending too much time getting OpenCV to install properly.

Getting started

Creating the Docker image

• Install Docker on your OS - we'll leave those details up to you to figure out.
• Clone this repository and cd to it

  # git clone https://github.com/dorontal/docker-opencv-python
  # cd docker-opencv-python
  

• Run

  # ./build_image.sh
  

Running inside a container with the opencv-python image

• Run

  # docker run -e DISPLAY=:0 -v /tmp/.X11-unix:/tmp/.X11-unix \
      -it opencv-python bash
  

  and you'll be in a bash shell that can pop up x windows (e.g. try xterm).

• Here's an example where you share the contents of a directory on the host (<HOST DIR PATH>) on a directory inside your container (<CONTAINER PATH>); also sharing the video device /dev/video0 from host to container, as well as sharing the X11 socket for display:

  # docker run -e DISPLAY=:0 -v /tmp/.X11-unix:/tmp/.X11-unix \
      --mount type=bind,source=<HOST DIR PATH>,target=<CONTAINER PATH> \
      --device=/dev/video0 -it opencv-python bash
  

  Notes:

  • <HOST DIR PATH> is the directory containing the code you want to run, on the host machine - your regular development environemnt available before running the docker container
  • <CONTAINER PATH> is a directory in the container that will end up containing the contents of <HOST DIR PATH> in a read/write fashion
  • --device=/dev/video0 makes the host device /dev/video0 available on the container (as /dev/video0)
  • -v /tmp/.X11-unix:/tmp/.X11-unix makes x socket on host replicated on container