Researchers at the Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory and Microsoft have developed an algorithm that finds hidden links between paintings at New York City's Metropolitan Museum of Art and the Netherlands' Rijksmuseum.
The MosAIc system uses deep networks to find paired or analogous works from different cultures, artists, and media.
The team employed a novel image-search algorithm to uncover the closest match to a specific artist or culture, and retrieve the most semantically similar image.
To find analogous images from different cultures, the researchers used a conditional KNN tree that clusters similar images together, starting from the "trunk" and tracing the most promising "branch" until they reach the closest image.
MosAIC differs from similar tools because it requires only a single image to find matches, and unearths connections in whatever culture or media the user is interested in, rather than in paths.
From MIT CSAIL
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA