Tool to run image annotation campaigns on Wikimedia Commons, created to gather data for machine learning tasks. This tool is based on Wikidata Image Positions and licensed under AGPL-3.0. For the changes made see the git history.

You can also run the tool locally, which is much more convenient for development.

git clone <repo>
cd tool-wd-image-positions
pipenv install
pipenv shell
export FLASK_APP=app.py
export FLASK_ENV=development
flask run

On Wikimedia Toolforge, this tool runs under the image-annotator tool name. Source code resides in ~/www/python/src/, a virtual environment is set up in ~/www/python/venv/, logs end up in ~/uwsgi.log.

Make sure to add / update config.yaml.

If the web service is not running for some reason, run the following command:

webservice --backend=kubernetes python3.7 start

If it’s acting up, try the same command with restart instead of start.

To update the service, run the following commands after becoming the tool account:

webservice --backend=kubernetes python3.7 shell
source ~/www/python/venv/bin/activate
cd ~/www/python/src
git fetch
git diff @ @{u} # inspect changes
git merge --ff-only @{u}
pip3 install -r requirements.txt
webservice --backend=kubernetes python3.7 restart

The code in this repository is released under the AGPL v3, as provided in the LICENSE file.