How can digital tools assist researchers in the study of works of art? How can such tools help the general public explore, discover and connect to our cultural and artistic past?
The digitization of millions of artworks by museums, libraries and cultural institution brings unprecedented access and opportunity to engage with collections. It also poses new challenges in the search, retrieval, and analysis of visual material. Currently, online databases of art can only be searched through keywords, or tags, namely the textual metadata of the original object. Textual queries, however, are not designed to search visual information which has not been indexed – they cannot for instance find similar shapes, forms, or motifs. This is why a new tool for searching art through its visual attributes is needed.
The Replica project led by the DHLAB aims to build the first search engine designed specifically for the search and exploration of artistic collections (including paintings, drawings, engravings, sculpture and photography). This employs the latest state-of-the art artificial intelligence techniques, such as Deep Learning and Convolutional Neural Networks, for the search and display of information.
In partnership with the Giorgio Cini Foundation in Venice and Factum Arte in Madrid, the Replica project aims to digitize roughly one million artistic reproductions using these images to populate the new search engine and as the basis for new art historical inquiries.
Adam Lowe and Factum Arte have been collaborating with the Fondazione Giorgio Cini for a long time and have now built a specially designed circular scanning system, capable of recording 12 A3 sized double sided images at 400DPI per minute. The paired images are automatically downloaded with metadata tags and archived. The scanner is also equipped with a system of automatic object detection, a controlled light source and instant visualization of the images as they are being recorded.
Replica opens new avenues for exploring and engaging with artistic collections. It assists art historians in the study of iconography, the transmission of forms, styles and patterns, while engaging larger audiences in a process of learning and discovery.
Interface Demo
Keywords
Visual Pattern Discovery; Art History; Photolibrary; Deep Learning; Visual Similarities
People
Isabella di Lenardo / Project Head, Collaboratrice Scientifique, EPFL
Benoit Seguin / Ph.D. Student, EPFL
Frédéric Kaplan / Director DHLAB, EPFL
Carlotta Striolo / Intern, DHLAB – EPFL
Lia Costiner / Post Doc, DHLAB – EPFL
Other Institutions and Persons involved:
The Replica 360 Recto/Verso Scanner was a development designed and constructed by Factum Arte supported by the Factum foundation.
Adam Lowe / Director
Enrique Esteban / Engineering and software Developer
Dwight Perry / Concept design
Pedro Miró Infante / Production Manager
Istituto di Storia dell’Arte, Fondazione Giorgio Cini, Venezia
Andrea Barbon / Project Manager, Fondazione Giorgio Cini
Simone Guerriero / Scientific Advisor, Fondazione Giorgio Cini
Monica Bassanello / Scientific Advisor, Fondazione Giorgio Cini
Ilaria Turetta / Scientific Advisor, Fondazione Giorgio Cini
Remko Bigai / IT management and development
Alice Vivian / collaborator
Erica Mariani / collaborator
Jacopo Scarpa / collaborator
Ginevra Gioa / collaborator
Irene Marcuz / collaborator
Tania Mio Bertolo / collaborator
Costanza Blaskovic / collaborator
Gaspare Frassetto / collaborator
Martina Lughi / collaborator
Andrea Del Zoppo / collaborator
Davide Polloni / collaborator
Giorgia Chiminazzo / collaborator
Laura Brunello / collaborator