WikiVisage

Active learning facial recognition for Wikimedia Commons. Train an ML model to recognize specific people and automatically add P180 (depicts) Structured Data to matching images.

Log in with Wikimedia

How it works

1. Create a Project

Enter a Wikidata Q-ID and Commons category. The system scans all images, detects faces with bounding boxes, and bootstraps from existing P180 depicts claims.

2. Classify Faces

Review detected faces one by one with Yes/No buttons. After enough examples, the model classifies remaining faces automatically using face encodings.

3. Review & Edit

Check model classifications. Approve matches, reject false positives, redraw bounding boxes, or manually add missed faces. Each face is compared independently.

4. Send to Commons

Send approved matches to Commons as P180 depicts claims. Rejected bootstrap faces queue P180 removal, fixing incorrect existing data.