2015 – 2018
This project gave rise to ARCHiVe, marking the first collaborative effort among the three partner organizations (Fondazione Giorgio Cini, Factum Foundation, and EPFL DHLAB). It tested the synergy of these entities for the first time, setting an ambitious goal achievable only through the collective effort of all involved.
The Project
The Fondazione Giorgio Cini preserves an extremely rich iconographic photo library, encompassing local and international art history, architecture, urbanism, and culture. It comprises approximately 730,000 photographic positives stored in the photo libraries of the Institute of Art History.
The Replica project addresses two crucial issues:
- How to rapidly digitise this vast number of documents while preserving the original heritage and ensuring optimal technological accuracy.
- How to make the resulting database quickly searchable without the necessity of relying solely on textual search methods.
To meet the first of these needs, Factum has specifically designed a circular scanner (Replica 360 recto/verso) for digitising the Historic Photo Library in the most efficient way possible. Additionally, Factum has equipped Replica with software that allows for the storage of data and metadata (such as the file and its archival position, ID number, and other essential details) during the digitisation process. In this manner, Fondazione Giorgio Cini has initiated the creation of an extensive database of images and information. This substantial volume of data is being stored and analysed by the Digital Humanities Lab at the École Polytechnique Fédérale de Lausanne.
In order to develop a geometrically-based search engine for this database, DHLAB team conducted their research focusing on recognising patterns of similarities in the pictures of the Historic Photo Library. DHLAB incorporated one of the most advanced technologies in artificial intelligence, known as ‘deep learning,’ into the process. By leveraging a vast network of artificial neurons, a capability only recently achievable in image processing, various forms of visual entities can be simultaneously analysed in a unified approach.
In the realm of machine learning experimentation, DHLAB also devised an automatic segmentation process that separates text and images, subsequently annotating them.
In fact, the Historic Photo Library comprises thousands of photographic positives dating from the beginning of the 20th century until the 90s, affixed to cardboards, with each displaying additional information such as the subject, location, date, and the artist of the depicted artwork.
objectives
The aim of the project is the digitisation and enhancement of Fondazione Giorgio Cini Photo Libraries through innovative mass recording and analysis techniques.
The overarching objective is to enable the online availability of exceptionally large archives. The enhanced computational capacity, tailored to handle specific data, facilitates user searches at various, increasingly refined levels.
Methodologies
Results
The aim of Fondazione Giorgio Cini is to proceed with the project by digitising the additional photo libraries stored in the Biblioteca Manica Lunga and to make this heritage accessible online.