cvbuilder

OpenCV Hands-Free

Unofficial OpenCV builder for Python.

This package aims at building OpenCV 4.5.4 with Python bindings from the official sources. It provides a simple command line interface for starting the process of downloading the official sources, configuring the build dependencies, compiling and installing the resulting CV2 shared object within a virtual environment.

In contrast to opencv-python it will not provide any wheels and therefore the installation / build process will be by far slower (depending on the actual system performance).

IMPORTANT NOTE

Depending on the usage and system dependencies, the on-the-fly build output can contain video and GUI functionality and the contrib package.

Features

• builds against Python >= 3.6
• runs and builds inside virtual environment
• includes video support
• supports OpenCV check (import, build information)
• compiles with many flags enabled (which???)

Supported OS

• Debian Jessie, Stretch
• Ubuntu 18.04
• LinuxMint 18.2
• Manjaro
• Arch
• Raspbian 9 (Stretch)

Supported Python runtimes

• CPython 3.6, 3.7, 3.8, 3.9
• PyPy (not yet)

Note: PyPy (7.3.0) is not supported yet due to missing path variables in module sysconfig.

Installation

• Use a Python's virtual environment or even better add the package via poetry: poetry add cvbuilder
• Follow instructions below

Usage

After installing the package via pip / pipenv / poetry, you can manually invoke cvbuilder commands.

Install system dependencies

cvbuilder system --enable-gui --enable-video

Download, configure, build, install

The do-it-all command is:

cvbuilder build

or if you already downloaded the source zip files in the default temporary directory or need to rerun the process in a clean way:

cvbuilder build --clean

Custom

If you need to run the individual steps (i.e. for debugging) the following commands are provided.

Download sources

cvbuilder download

Generate make config

cvbuilder configure [--tmpdir XYZ]

Compile

cvbuilder make

Install

cvbuilder install

Dump

cvbuilder dump

Check

You can run the check command:

cvbuilder check

in order to check whether the build process and installation worked.

Running OpenCV

Now everything should be up and running and you should be able to work with OpenCV:

• Start a python REPL: poetry run python
• Import the CV package: import cv2
• Read OpenCV documentation