Visually Digging into Museum Data

How can a museum community with an active online user base collaborate in fundamental computer science research to the benefit of both researchers and museum audiences?  How can the museum’s digital assets inform fundamental computer science research questions? How can the computer science research and algorithm development be translated into user tools for the museum community?

This demonstration will present the work of Michigan State University’s MATRIX: Center for Humane Arts, Letters and Social Sciences Online and Michigan State University Museum on the Quilt Index through the collaborative research project: Digging into Image Data to Answer Authorship Related Questions.  Three project teams of computer scientists, museum professionals, and humanities scholars from Michigan State University, University of Illinois at Urbana-Champaign, and University of Sheffield (UK) have been exploring authorship studies of visual arts through computational image analyses. The three project teams are utilizing three datasets of visual works—15th-century manuscripts, 17th and 18th century maps, and 19th to 21st-century quilts—to investigate what might be revealed about the authors and their artistic lineages, and to test how algorithms can be applied across diverse data sets of images.

The computer science approach to this problem aims to discover what salient characteristics that differentiate artists can be applied to enable statistical learning about individual and collective authorship, given a set of 2D images of historical artifacts with known authors. The objective of this effort is to develop automated analyses to process specific salient characteristics on local data sets, hone those processes to provide the results at a much higher level of confidence than previously has been feasible, and test those processes across 1) diverse and 2) large sets of images.  

This paper and demonstration will analyze the effort’s collaborative methods for project development, the research challenges, and the potential for museum collaboration, including lessons learned about collaborating on computer science research and ways to apply such research to develop public user applications for museum website visitors.