Digging into Image Data to Answer Authorship Related Questions (DID-ARQ) seeks to explore authorship studies of visual arts through the use of Computer Vision. In the past, authorship has been explored in terms of attributions, typically of either individual masterpieces or small collections of art from the same period, location, or school. Due to these localized strategies of exploration and research commonalities and shared characteristics are largely unexplored. In fact, it is rare to find discussions beyond a single discrete dataset. More significantly, to our knowledge, there have to date been no studies of image analyses targeting the problem of authorship applied to very large collections of images and evaluated in terms of accuracy over diverse datasets.
DID-ARQ investigates the accuracy and computational scalability of image analyses when applied to diverse collections of image data. While identifying distinct characteristics of artists is time-consuming for individual researchers using traditional methodologies, computer-assisted techniques can help humanists discover salient characteristics and increase the reliability of those findings over a large-volume corpus of digitized images. Computer-assisted techniques can provide an initial bridge from the low-level images features, such as colors or pixels, to higher-level semantic concepts such as brush strokes, compositions, or patterns. This effort will utilize three datasets of visual works: 15th-century manuscripts, 17th and 18th-century maps, and 19th and 20th-century quilts to investigate what might be revealed about the authors and their artistic lineages by comparing manuscripts, maps, and quilts across four centuries. Based on the artistic, scientific or technological questions, DID-ARQ intends to formulate and address the problem of finding salient characteristics of artists from two-dimensional (2D) images of historical artifacts. Given a set of 2D images of historical artifacts with known authors, our project aims to discover what salient characteristics make an artist different from others, and then to enable automated classification of individual and collective authorship.
National Science Foundation – EAGER: Digging into Image Data to Answer Authorship Related Questions (IIS-1039385), 2010-2012
Kenton McHenry (PI)