About the project

The FAME (Facial Recognition as a Tool for Metadata Creation) project, supported by the Flemish Government and led by Meemo, explored the application of facial recognition technology to improve metadata generation for digital heritage collections. Meemo has a large archive of digitized pictures of different theater shows, live performances, and cultural events spanning multiple decades. Most of these images are not sufficiently annotated and do not contain any information on who is in the images. Therefore, machine learning tools were used to detect the persons visible on the images and provide additional metadata. Using a known dataset of actor portfolios and facial recognition tools, we were able to automatically detect who is depicted on the images and improve the collection quality.

Project Scope and Achievements

• Objective: To enhance the accessibility and searchability of cultural heritage collections through (semi-)automated facial recognition.
• Pilot Partners:
  • Kunstenpunt (Flanders Arts Institute) – Performing artists
  • KOERS Museum of Cycle Racing – Professional cyclists
  • ADVN & Flemish Parliament Archive – Politicians and campaigners
• Scale of the Project:
  • Over 150,000 photos and videos analyzed
  • 182,202 validation labels assigned
  • 78,440 correctly identified faces
  • 1,694 unique individuals successfully recognized

IDLab role

1. Development of Facial Recognition Pipeline

• Implemented and optimized an open-source facial recognition tool
• Setup labeling tool in collaboration with archival organizations for evaluation
• Achieved high accuracy in face detection and matching (93.6 %)

2. Integration with Metadata Systems

• Enabled seamless linking of identified faces to archive management platforms
• Facilitated structured metadata creation and enrichment

4. Collaboration & Knowledge Sharing

• Worked alongside Meemoo and cultural partners to refine the tool
• Contributed to FAME study days, focusing on ethical and legal aspects

Impact & Future Prospects

The project demonstrated that semi-automated facial recognition significantly aids metadata creation. While challenges remain, such as manual validation and reference set setup, FAME’s insights are now influencing:

• The GIVE metadata project, which aims to further explore facial recognition applications.
• Future proof-of-concepts for a large-scale meemoo archive system.

By addressing both technological and ethical concerns, FAME has set the stage for broader adoption of facial recognition in cultural heritage archiving.

Through these efforts, IDLab significantly contributed to the advancement of automated metadata creation, demonstrating the practical applications of facial recognition technology in cultural heritage preservation.