Installation

Requirements

The plugin’s logic is mainly based on a well-known machine learning toolkit: scikit-learn. To work, the plugin needs it and it’s quite a challenge to install it on QGIS for Windows, because QGIS uses its own Python interpreter and doesn’t make it easy to use packages manager (pip).

To make it easier for Windows end-users, we did our best to embed the dependencies within the released version of the plugin, inside the embedded_external_libs folder.

Technically, the plugin tries to:

  1. import Scikit-learn from the Python interpreter used by QGIS (system on Linux, specific on Windows)

  2. add embedded_external_libs to the PYTHONPATH and then import Scikit-Learn from it

BUT there are some caveats because

Linux

Mainly tested on Ubuntu 20.04.

  • 3.16 =< QGIS =< 3.22

  • Python 3.8+

  • dependencies listed in requirements/embedded.txt. On Linux, you need to install them by yourself. Typically, on Ubuntu:

# using the requirements file if you have the plugin repository
python3 -m pip install --no-deps -U -r requirements/embedded.txt
# or directly - refer to the file to be sure of using the right updated pinned versions
python3 -m pip install --no-deps -U "joblib==1.1.*" "scikit-learn==1.0.*" "threadpoolctl>=2,<3"

Important

Remember to look at the requirements/embedded.txt file to make sure you are using the right versions. The above command is just an example, a pattern.

Windows

Mainly tested on Windows 10.0.19044.

  • 3.16 =< QGIS =< 3.22

  • Python 3.9 (strictly)

Use it with a different Python version (manual install)

  1. Launch the OSGeo4W Shell. Look for it in your Windows Search or directly into your QGIS install folder:

    • installed with the all inclusive standalone .msi: C:\Program Files\QGIS

    • installed with the OSGeo4W with default settings: C:\OSGeo4W

  2. Run:

python-qgis-ltr -m pip install -U pip
python-qgis-ltr -m pip install -U setuptools wheel
python-qgis-ltr -m pip install -U "joblib==1.1.*" "scikit-learn==1.0.*" "threadpoolctl>=2,<3"

Important

Remember to look at the requirements/embedded.txt file to make sure you are using the right versions. The above command is just an example, a pattern.

Stable version (recomended)

This plugin is published on the official QGIS plugins repository: https://plugins.qgis.org/plugins/RoofClassify/.

Go to Plugins -> Manage and Install Plugins, look for the plugin and install it.

Beta versions released

Enable experimental extensions in the QGIS plugins manager settings panel.

Alternate plugins repository

During the release workflow (see Packaging and deployment), the plugin is packaged and released on the official QGIS plugins repository but also as a GitHub Release, right aside the repository.

You can add the following URL to the QGIS extensions repositories, in the Settings tab of the Plugins Manager window (see official documentation):

https://github.com/DINFO-UniFI/RoofClassify/releases/latest/download/plugins.xml

Warning

As GitHub Releases publishes only stable releases through its latest API, only not tagged releases as pre-release are available.