Debian-OpenCV2

This docker image provides latest stable Debian, with OpenCV 2.4 and Python 2.7 installed, plus some of the popular Python packages, plus several ways to use graphics.

Installing Docker

Ubuntu

sudo apt install docker.io
sudo usermod -aG docker $(whoami)

Until you re-login, you will have to prefix all docker commands with sudo.

Windows

Download "Docker Toolbox" from https://www.docker.com/products/docker-toolbox, install it, run "Docker Quick-start Terminal", and let it do its magic.

You can also try "Docker for Windows", which sounds like a better option, but is unsupported on my Windows machine.

Working with this image

First, do docker pull al42and/debian-opencv2:latest to get an image (it's about 1GiB).

If you want to bind host directory within container, use docker run -v $(pwd):/app option or the like. This might cause some problems on Linux due to different user IDs.

Using PyCharm Professional

Open PyCharm, go to Preferences (or File -> Settings) -> Project:yours -> Project Interpreter, and add new Remote Interpreter.

Using Jupyter

docker run -p 9999:9999 -dt al42and/debian-opencv2 run_jupyter

And then open http://localhost:9999/ in your browser

Jupyter and X in Linux (probably MacOS too)

With some minor effort (and a bunch of X11 magic) we can do the following to pass X socket inside container:

XSOCK=/tmp/.X11-unix
XAUTH=/tmp/.docker.xauth
xauth nlist :0 | sed -e 's/^..../ffff/' | xauth -f $XAUTH nmerge -
docker run -p 9999:9999 -dt -v $XSOCK:$XSOCK -v $XAUTH:$XAUTH -e XAUTHORITY=$XAUTH -e DISPLAY=$DISPLAY al42and/debian-opencv2 run_jupyter

And now we can watch OpenCV videos or use Matplotlib animations from Jupyter notebook!

Using SSH X-forwarding

docker run -p 2222:22 -dt al42and/debian-opencv2 run_ssh

This way, you have this container running in background. Now, if you're on Linux or Mac, just do ssh -X root@localhost -p2222, with password being 1234.

If you're on Windows, things get more complicated. First, run docker-machine ip default in Docker console to get an IP of a virtual machine Docker uses. Then install Xming and use its XLaunch tool to connect to this IP, port 2222, username root and password 1234.

Jupyter and X in Windows

Surprisingly, this paragraph is pretty short. First, install Xming and run container as:

docker run -p 2222:22 -dt al42and/debian-opencv2 run_ssh

Then download file tools/run_jupyter.xlaunch from this repository, right-click it and choose edit. XLaunch should open. Accept all default choices, but change host IP to the one returned by docker-machine ip default.

After several seconds it should launch X-server on your machine, and Jupyter in container, with SSH connecting them. Now, open http://localhost:9999/ and rejoice!

Jupyter and X in Linux/MacOS

On Linux it's more efficient to use Unix sockets to connect to X server, but you can go ssh-way if you want:

docker run -p 2222:22 -dt al42and/debian-opencv2 run_ssh
ssh root@localhost -p2222 -L9999:localhost:9999 -Xf run_jupyter &>/dev/null

It start background ssh connection to container, with port 9999 and X forwarding.

Using VNC

docker run -p 5900:5900 -dt al42and/debian-opencv2 run_vnc

Then connect to localhost:5900 using your favorite VNC client. The password is 1234.

Building

docker build -t al42and/debian-opencv2 --no-cache .

Known issues:

Mac OS - Certificate path doesn't exist:

See https://stackoverflow.com/questions/38285282/where-is-the-certificates-folder-for-docker-beta-for-mac

When doing import cv2 you get libdc1394 error: Failed to initialize libdc1394:

Ignore it, there's nothing to worry about

al42and/docker-debian-opencv2