A paper released earlier this month by a group of Rutgers University researchers applies computer vision and machine learning to the question of artistic influence. For their study, Babak Saleh, Kanako Abe, Ravneet Singh Arora, and Ahmed Elgammal used algorithmic analysis to discern the recurrence of visual patterns between artists. Though the authors explain that the use of algorithms for classifying artworks is not new, the application of mathematical principles to the question of influence remains novel. “Measuring influence is a very difficult task because of the broad criteria for what influence between artists can mean,” they write.
Unlike the two major initiatives extant to give order to aesthetic fields, Artsy’s Art Genome Project and the earlier Music Genome Project by Pandora, the study’s methodology does not rely on an extensive taxonomy of verbal descriptors assigned by human experts to evaluate similarities between artworks. The dataset, encompassing 1,710 paintings by 66 artists, is relatively small, but the researchers’ algorithmic analysis was able to independently confirm various established relationships of influence (e.g. Delacroix on Bazille, Munch on Beckmann, Degas on Caillebotte, etc.) as well as identify some interesting proximities between artists across movements.