The new algorithm developed by his researchers WITH, "MosAIc", Finds indistinguishable similarities between works of art, scanning them pictures.
The new system works in Metropolitan Museum of Art and Rijksmuseum of Άμστερνταμ. The MosAIc algorithm scans the artwork and then uses it deeply networks to find similarities in pieces from cultures, artists and media that may not have been noticed, according to a lab blog post Computer Science & Artificial Intelligence (CSAIL) of MIT.
But let's see how it works: the user enters an image and MosAIc finds similar artwork. In one example, the algorithm linked the "The Martyrdom of Saint Serapion » of Francisco de Zurbarán, and "The Threatened Swan » of Jan Asselijn.
«These two artists did not correspond, nor have they met during their lifetime, but their paintings revealed a rich, hidden structure that governs both of their works., "Said the MIT CSAIL student, Mark Hamilton, its lead author research.
A particularly difficult aspect of MosAIc's development was the creation of an algorithm that could find not only similarities in color or style, but also in meaning and subject matter, he said. Hamilton. The researchers examined a deep network of "activations" or features, for each image in the open access collections of both museums. The distance between the "activations" in the deep internet, was the way the researchers judged the similarity.
MosAIc researchers also used a new data structure search images called "KNN tree, Which groups the images into a tree-like structure. To find the closest match of an image, the algorithm starts from its "trunk" grouping and then follows the most promising "branch" until it finds the closest picture. The data structure improves on its own by allowing the tree to "prune" itself based on the characteristics of the image.
Hamilton said he hoped the work started at MosAIc could be extended to other areas, such as the humanities, social sciences and medicine.