Things change. Half an year ago, I was happy with running OpenCV on Python 2. But not anymore. As hard as it is for me to admit, it is time to switch to Python 3. So here is my supposedly easy to follow tutorial on setting up OpenCV on Python 3.
First of all, get the latest package listings from the repository. If you are on fedora, you can safely skip this. If you are on Debian, do the following.
$ sudo apt-get update
Lets make sure Python 3 and its dev libs are installed. If on Debian, run
$ sudo apt-get install python3 python3-dev
If you are on fedora,
$ sudo dnf install python3 python3-devel
Now, let us install the python package manager, more commonly called pip. On Debian, you would want to do this,
sudo apt-get install python3-pip
On fedora, you can do
$ sudo dnf install python3-pip
Now we can proceed to install VirtualEnv.
Virtualenv is a container inside which you can set a Python environment complete with libraries which is independent from the rest of the system. Setting up Virtualenv is worth it, as it allows us to maintain multiple versions of libraries parallely without conflicts.
On fedora, do
$ sudo dnf install python3-virtualenv
If you are on Debian, do
$ sudo apt-get install python3-venv
Once installed, go to someplace where you want to keep the files for your project. Create a directory for the same and initialize a virtualenv inside the directory. It can be done as follows
$ mkdir opencv
$ cd opencv
$ python3 -m venv .
To make a virtualenv work for us, we need to activate it. This is done by calling the activate script.
$ source bin/activate
Once this is properly done, your prompt should change to something that resembles
(opencv) $
That should do it. Let us install the libraries.
Now we address the elephant in the room. Let us setup OpenCV and some of the essential packages to make it work for us.
$ pip install opencv-contrib-python # the OpenCV package
$ pip install imutils # convinience functions
$ pip install matplotlib # easily plot the resultant images
$ pip install jupyter # easy, powerful programming environment
$ pip install numpy # for managing all those image arrays
If you want an easy to use programming environment for learning OpenCV, you just installed it. We are ofcourse, talking about Jupyter Notebook. To start, do
(opencv) $ jupyter notebook
As always, once you are done coding, you can quit by presssing “Ctrl + C”, then deactivate Virtualenv by running “deactivate” command
(opencv) $ deactivate