A team from MIT (CSAIL) and Microsoft developed an algorithm that can discover connections between pieces of art that belong to different artists and cultures.
It is not about the inspiration or influence that certain artists may have had on the works of others. Or if they share the same style or color. This AI goes much further, detecting hidden connections or coincidences between artists who did not even know of the existence of the other.
To illustrate the kind of connection they are referring to, Mark Hamilton, one of the creators of the algorithm, mentions two works of art that were paired in an exhibition at the National Museum in Amsterdam. The Martyrdom of Saint Serapion, an oil painting on canvas from 1628, by Francisco de Zurbarn, is on display. And The Threatened Swan, from 1650 by the painter Jan Asselijn.
While the artists never met and the works have different meanings, they found that they have some visual similarity. A very difficult match to establish. To obtain the same result, they used image retrieval technology in the development of the MosAIc algorithm, starting with the database of the National Museum of Amsterdam and the Metropolitan Museum of Art.
Using a single image as a reference you can search for connections taking into account different established criteria. This opens up a world of possibilities, since the algorithm can search among works of art of different styles, few and cultures. It will be like rediscovering works of art from these coincidences.
The MIT team explains in detail how they developed this algorithm and the potential for its dynamics in different areas in a PDF document.
