Face-Based Metadata - Doug Peterson Imaging Product Manager, Image Science, Product Evangelist, AI guru, and Photographer

Executive Overview

As Head of R&D at Digital Transitions, I led a pioneering consulting engagement to apply artificial intelligence to enrich the photographic archives of 24 Estonian cultural institutions. I designed and delivered an AI-powered facial recognition solution that achieved:

Cross-Collection Discovery: Identified ~300 individuals appearing across multiple institutional collections, connecting previously isolated materials
Research Enhancement: Revealed historical narratives by connecting images of individuals across time and context
Identification: Automatically labeled 343 notable historical figures

This project significantly increased the value of the digitized collection by enabling researchers to navigate by individuals, and secured our company's role in digitizing the remainder of the consortium's collections.

I am currently engaged in a similar project at the archives of the National Geographic Society.

Lorem ipsum

Lorem ipsum dolor sit amet, consectetur adipisicing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum.

Lorem ipsum

Client Challenge

The 24 Estonian cultural institutions faced significant challenges with their newly digitized collections:

Metadata Gap: Large volumes of digitized material lacked descriptive information
Resource Constraints: Limited staff capacity for manual identification of individuals
Institutional Silos: Potentially related materials remained disconnected across institutions
Untapped Potential: Valuable research connections remained hidden in unstructured visual data

Our Technical Approach

Discovery & Assessment

Collaborated with stakeholders representing all 24 institutions
Assessed technical and privacy requirements and developed specialized data handling protocols
Created processing pipeline to transform archival TIFFs into AI-ready images
Established clear success metrics and deliverables with client stakeholders

AI Implementation

Face Detection: Used AWS Rekognition to identify approximately 118,000 faces across 45,000 images
Vector Embedding: Used AWS Rekognition to create mathematical representation of each face
Custom Clustering: Developed network graph approach to identify individuals appearing multiple times
Celebrity Recognition: Applied AWS Rekognition with historical filtering to identify notable individuals
Quality Control: Created visual review tools for institutional validation of results

Results Delivery

Network Visualizations: Generated visual representations showing relationships between facial clusters
Photo Mosaic: Created Estonian Coat of Arms composed of 70,000 detected faces
Frequency Analysis: Identified individuals appearing multiple times across collections
Cross-Collection Mapping: Revealed connections between formal portraits, candid appearances, and even artwork
Presentation: Conducted live webinar with staff from across the consortium's institutions

Lorem ipsum

Value Delivered

Cross-Collection Intelligence: Connected individuals across institutions, contexts, and media types (even identifying subjects in paintings)
Enhanced Discovery: Enabled navigation by depicted individuals, identified 343 notable figures, and prioritized high-frequency subjects
Research Catalyst: Revealed previously invisible connections, enabling new historical narratives and exhibition opportunities

"DT's PixelFlow AI was successfully able to detect the same person in different ages and positions across many collections. The best example of successful recognition was exhibited in the case of actors who wear makeup and present themselves in different poses. As a tool for archivists, successful recognition helps to save a lot of time during the comparative description of images (making) the process of research much more convenient." – Aap Tepper, Senior Specialist, Film Archives, The National Archives of Estonia

Lorem ipsum

Success Factors

This consulting engagement succeeded by:

Leveraging Commercial AI Services: Using AWS Rekognition as a foundation while adding custom enhancements
Developing Specialized Methodologies: Creating custom clustering algorithms for historical photography
Addressing Cross-Institutional Needs: Providing solutions beyond what any single institution could achieve
Creating Tangible Visualizations: Making complex connections visible and enabling visual quality control
Cross-Domain Knowledge: Combining AI capabilities with cultural heritage expertise

The Estonia AI project demonstrates my ability to deliver high-value consulting services that unlock new research potential from digitized cultural heritage collections while building lasting client relationships.

Lorem ipsum

Watch Presentation

I presented this project live to a hybrid in-person/remote audience at the 2022 DT Roundtable. The relevant section starts at around 43:30.