Skip to end of metadata
Go to start of metadata

Our next Research IT Reading Group topic will be: Digital Art History, Th 24 September / noon / 200C Warren Hall.

When: Thursday, September 24th from noon - 1pm

Where: 200C Warren Hall, 2195 Hearst St (see building access instructions on parent page).

Event format: The reading group is a brown bag lunch (bring your own) with a short <20 min talk followed by ~40 min group discussion.

Presenter: Elizabeth Honig, Associate Professor, History of Art

Elizabeth Honig will discuss how digital tools are changing the study of art history and offering new opportunities for collaboration. She will discuss two of her current projects:

  • developing an open source platform for creating catalogues raisonnés (resources that meticulously detail the oeuvre of a particular artist and provide evidence of the provenance and attribution of the artist’s works) and distributing it as a Drupal module (funded by the Digital Humanities at Berkeley program, a partnership between Research IT and the Dean of Arts and Humanities).

  • collaborating with Eric E. Monson,of Duke University’s Visualization & Interactive Systems Group on NSF-funded research in computer vision and machine learning

Please review the following prior to our 9/24 meeting:


Presenting: Elizabeth Honig, Art History


Aron Roberts, Research IT
Camille Villa, Research IT
Chris Hoffman, Research IT
Jason Christopher, Research IT
Laura Sheffler
Mary Elings, Bancroft Library
Patrick Schmitz, Research IT
Raymond Yee, D-Lab
Rick Jaffe, Research IT
Ronald Sprouse, Linguistics
Scott Peterson, Doe Library
Steve Masover, Research IT
Steven Carrier, School of Education
Zac Bleemer, Economics

Presentation (Elizabeth Honig):

Art History doesn't think of itself as data oriented scholarship.

What is data in Art History: metadata is object dimensions, materials, date, where object is located, provenance, history of ownership.

Jan Breugel the Elder. Flemish, worked in Italy, painted a large range of different kinds of work. Different sizes, materials, subjects. Repetition within his works. Different supports (e.g., copper, panel). Thousands of works related to JB's studio, related to each other more and less. Value for a good work: $10M; a knockoff in the studio, still worth $100K. Stakes are high, so scholarship (particularly about attribution) is skewed by financial opportunity ... and 'market' on JB (esp. Catalog Raisonné) is 'cornered' by a particular individual whose work is less than accurate, and owners of objects don't want to release permission to anyone else to catalog their work because it might be declared inauthentic.

Elizabeth attempted to make her own catalog in order to develop a more accurate set of metadata about JB's work.

Web site to engage others in contributing and correcting data in the catalog. Interestingly, many people (distributed) control small amounts of information about particular works, e.g., curators at museums that hold JB's work. Working with Comp Sci person from Davis, got CITRIS grant to build a web site to facilitate contribution from a distributed set of individuals.

Gathered data for ~3 weeks on trip to Europe w/ cameras and copy-stands, harvested data from two archives in The Netherlands and one in the Netherlands.

With this data, built web pages atop a MediaWiki platform. Should have refined the data first. Was tempted by the opportunity to get others to contribute. Recruited contributors at conferences etc., gave away lots of swag. [Turns out that very few actually contributed a great deal. That's a problem that remains to be solved.]

Pages included metadata about the painting; also related works to the subject of the page; and -- behind each page -- discussion among contributors. Search tools, including advanced search on a number of criteria of particular interest to art historians. Coming soon: display of paintings at relative scale to each other when shown on a web page. Images are tagged to identify objects in the painting (e.g., temple, waterfall). Also browsing by genre (e.g., religious).

What does "related" mean -- an art historical question that the site gives tools to resolve (and we expect to be able to do better as tools evolve in collaboration with computer scientists)? Techniques of production is an important aspect (e.g., how were components of images reused in multiple works created by artists and artisans in JB's studio).

Comparison tool (algorithmic): very helpful to have detail of two images easily presented on a digital screen, rather than move one's gaze between details in two much larger images. Crop a transparency, ability to flip or tilt, then can drag atop another painting and compare directly. By seeing which elements and which relationships between elements are identical and different one can learn a great deal about how the different images were produced from, say, a single composition drawing or a drawing of some single element within the composition. Noteworthy that these are aids to human identified similarities and differences.

What when the group of materials needing examination is large or diversely divergent. So: machine learning. NSF grant, work with folks at Duke: training algorithms to pick out related details between multiple images. EH contribution is humanly-judged relationships; the computers are then trained to find likeness based on training. In the course of this, Duke collaborators wanted metadata in a database, but the MediaWiki site required metadata to be harvested from individual objects. With Quinn Dombrowski's help/advice, metadata is being extracted into spreadsheets of horrifying magnitude for migration to a database-driven site built on Drupal. This turns out to be a very tedious and complex process.

Once the site is migrated and the data rationalized we'll develop a platform that is well-structured and populated with appropriate tools for application to other art history projects than just the JB project.

Overall goal: to make data widely available -- in addition to interpretive monograph, also written by EH -- as an incentive to further research. The platform is meant to encourage similar data sharing strategies for other art historians.


Patrick: Have others approched w/ interest in a different collection of objects
EH: Have used Mellon funding to explore interests and needs of >30 scholars (including an Edward Hopper scholar). Different ways of thinking about things or different interests -- some surprising -- e.g., catalogs of contemporary art use different media, questions of storage are quite different, exhibition history is quite important for contemporary art (whereas it is not for 17th century art). Very good to learn this early because it helped us to modify data and function design.

Chris: Museum Informatics program, way at the beginning of EH's project someone contacted MI on EH's behalf, but there wasn't an obvious way to help ... terrific to see that EH found collaborators and funding and kept it going.

Mary: Availability of data set?
EH: Can't download the whole thing, but can explore, can save one's own result sets.
Mary: Will a researcher be able to cite if they publish based on these data
EH: They will be able to.
Mary: Who did EH reach out to as tool / site was being designed
EH: Many art historians, also curators. Code Art (conference); next year in Madrid, for Bosch exhibition. One interesting discovery is that European curators are too busy to even contemplate

Raymond: Would you mind if someone downloaded the data in whole
EH: No.
Mary: People could do different kinds of research and manipulation of these data.
EH: Can't imagine why someone would want to do that
<all>: Yes, people will.
Chris: Develop a licensing / citation policy.
EH: Once my book is out next year, my interpretive work on these data will be done. Next book is Peter Breugel. 500-700 visitors/month now, will increase dramatically once we have PB data available.

Aron: Cataloging tools, e.g., CollectionSpace and its competitors. What features beyond image investigation are useful to Art Historians?
EH: Search, relationships between metadata-categorized aspects of work.
Aron: Forensic aspects of image comparison and ML seem quite significant.
EH: Dealers love this. But it's shady business, so we tend to steer clear of this kind of work -- only with trusted dealers ... but it's dangerous to become obligated to dealers whose business includes sale of less authentic or inauthentic work.
Aron: licensing / broad use opportunity -- market?
EH: Worth a thought, but seems to go counter to EH interest in open access.
Aron: a network of sites focusing on different artists
EH: In looking for a partner to develop sites, candidate European partners wanted to insist on multiple artists as focus of a build-out ... but that was too daunting given how much of EH's life focus on the Breugels has taken
Chris: would be very happy to provide advice about architecture ...
EH: Quinn ...
Chris: So Quinn and Drupal framework then
EH: Yes.

Raymond: What would it be to load up open source data from Wikimedia Commons, for example -- many artists' images?
EH: No reason for that. The system offers more functionality than one would need.
Raymond: Image comparison...
EH: Interesting idea. Also: scale gallery function, as on a museum wall (see:
Raymond: visualization equipment (glasses) for immersive experience in such a virtualized gallery

Rick: Scholarship about JB's studio
EH: I have about half a chapter in my forthcoming book about the studio ... no one else has done work like this to-date. There are some documentary materials.


  • No labels