Florence + The Machine: A Computational Approach to Florentine Liturgical Manuscript Illuminations from the Late Trecento (thesis)
Washington and Lee University -- Honors in Art History
Neural networks (Computer science)
Illumination of books and manuscripts
MetadataShow full item record
Thesis; [FULL-TEXT WILL BE AVAILABLE FOLLOWING A 3-YEAR EMBARGO]Elyssa McMaster is a member of the Class of 2022 of Washington and Lee University.This thesis investigates early methods to use artificial intelligence to sort images of medieval Italian manuscript illuminations by workshop from 1360 to 1400. Sorted images were passed through a 2-dimensional convolutional neural network to train a computer to discern the authorship of an unlabeled image. The network reached a 78% rate of accuracy. Though the network did not achieve high enough accuracy to be consistently reliable, preliminary results indicate that this project has produced the first step to continue objective connoisseurial work on this dataset. This project indicates that Jacopo di Cione, Niccolò di Pietro Gerini, and several of their assistants were the masters responsible for a body of manuscript illuminations previously attributed to a single figure named Don Silvestro dei Gherarducci.