| Platform | Status |
|---|---|
| Windows | |
| Manylinux |  |
Unofficial OpenCV packages for Python.
Work in progress!
The aim of this repository is to provide means to package each new OpenCV release for the most used Python versions and platforms.
At the same time it allows anyone to build a custom version of OpenCV for any Python version: just fork this repo and modify the build files and scripts to fit your needs.
- Installation of OpenCV for Python is pretty hideous:
- Download OpenCV
- Find cv2.pyd from the package
- If it exists, copy it to the root of Python site-packages
- If it does not exist for some reason for your setup, you have to setup the build environment and compile it manually
- Try to import cv2 and hope it works
- Everyone should be able to install OpenCV (or any package for that matter) with pip with a single command without building anything
- Python wheels are nice, we should use them more
The project is structured like a normal Python package with a standard setup.py file. The build process is as follows (see appveyor.yml):
- Checkout repository and submodules
- OpenCV is included as submodule and the version is updated manually when a new OpenCV release has been made
- Find OpenCV version from the sources
- Upgrade pip and install numpy for each Python version
- Build OpenCV
- tests are disabled, otherwise build time increases too much
- Copy each
.pyd/.sofile to cv2 folder of this project and generate wheel - Install the generated wheels for each Python version
- Test that the Python versions can import them
- TO DO: upload the wheels to PyPi
Currently the find_version.py file parses OpenCV version information from the OpenCV sources. OpenCV depends on numpy, so setup.py checks the numpy version also with the help of pip.
As described earlier, for example the .pyd file on Windows is normally copied to site-packages. To avoid polluting the root folder the __init__.py file in cv2 folder handles the import logic correctly by importing the actual .pyd module and replacing the imported cv2 package in sys.modudes with the cv2 module to retain backward compatibility.
Linux wheels are built using manylinux. These wheels should work out of the box for most of the distros out there since they are built against an old version of glibc.
Currently the find_version.py script searches for the version information from OpenCV sources and appends also a revision number specific to this repository to the version string.
Opencv-python package (scripts in this repository) is available under MIT license. The OpenCV itself is available under 3-clause BSD License (LICENSE-3RD-PARTY.txt).
A release is made and uploaded to PyPI when a new tag is pushed to master branch. These tags differentiate packages (this repo might have modifications but OpenCV version stays same) and should be incremented sequentially. In practice, release version numbers look like this:
cv_major.cv_minor.cv_revision.package_revision e.g. 3.1.0.0
Every commit to the master branch of this repo will be built. Possible build artifacts use local version identifiers:
cv_major.cv_minor.cv_revision+git_hash_of_this_repo e.g. 3.1.0+14a8d39
These artifacts can't be and will not be uploaded to PyPI.
There's a build time limitation (AppVeyor open source builds may take max. 1 hour) which restricts the supported Python versions to two. As Python's 2.x releases are slowly approaching legacy state, 2.7.x releases will be the only supported Python 2 versions on Windows. On Python 3 side, builds will be run only for the latest release.
However, if you wan't to get some other versions, just fork this repo and change the dependencies.
Manylinux wheels are built for all the Python versions which are supported by the manylinux containers.
TODO