Blog

A late start to the blogging season
Sunday, October 19, 2014

Whew! What a start to the academic year, one so busy I fear I have neglected this important corner of the website.

A great DIG Cohort is active this fall. One will want to pay attention to this space for reports of their progress. Indeed, Cecilia Wichmann, one of its members, will post here tomorrow. I'll have a post next week offering a report back from my trip to San Antonio and the Mid-America College Art Association conference, where I will present on the Collaboratory. I look forward to the scholarly and professional exchanges there, including catching up with Juliet Wiersema, Maryland PhD alum and co-organizer of the panel.

It's a great time in the Collaboratory and I look forward to sharing with you, in addition to guest posts from others, our many exciting adventures.

Quint

 

There's something happening here. What it is ain't exactly clear.
Wednesday, March 19, 2014

With apologies to Buffalo Springfield.

Even though Spring Break holds forth for many of us this week, an intellectual row of sorts (some of us might contend that it is a bit of an unfair fight) erupting in today's pages of political wonkery (fivethirtyeight, New Republic) is compelling as a snapshot of some of the same paradigmatic shifting that is roiling the humanities. Nate Silver, doyen of a geek-set passionate about data-driven journalism, has just relaunched fivethirtyeight, his blog so named for the numbers of electors responsible for selecting/electing the President and Vice-President of the United States and a blog that has achieved remarkable prominence for the remarkable precision AND accuracy of the election predictions contained therein, especially for the last two presidential contests. The relaunch, a homecoming of sorts at ESPN, comes after Nate Silver published for a few years at the New York Times. One of the explanations put forth (but not, it should be noted, verified by Nate Silver himself) is that he did not get along with the journalists at the Times. This could be true; Nate Silver does not hold opinion columnists in high regard, and it would not be surprising to learn that many of the columnists at the Times felt equally wary of Silver.

Coming on the heels of Silver's relaunch of fivethirtyeight is Leon Wieseltier's essay in today's New Republic, "The Emptiness of Data Journalism Nate Silver could learn a lot from those op-ed columnists he maligns." At its essence, Wieseltier's piece suggests that data-journalism of a type Silver advocates, replete with data visualizations, is soulless, lacking essential truth and avoiding a certain responsibility to stand for something (a very curious compaint). Perhaps Wieseltier objects to Silver's characterization of the work of opinion columnists, relying as many of them do on lazy anecdote (his view), as akin to that of the hedgehog, which knows only one thing, versus the fox, which sees and comes to know many things (after the suggestion by the Greek poet Archilochus that the fox knows many things and the hedgehog one big thing).

What interests me in this affair are less the specific arguments, although these are fascinating, than the remarkable amount of attention this spat attracted just today. Here's Charles Pierce in his Esquire Daily Politics Blog, Dylan Byers in Politico, Talking Points Memo airquoting Silver's "data intimidation", Paul Krugman plaintively asking that fivethirtyeight "tell me why the data matter."

That data-rich and data-focused journalism reaps such a bountiful harvest of criticism, positive and negative, at this time suggests to me a few things: the age of information visualization is upon us, as both practitioners of journalism and its consumers have matured in their appetites and capabilities for such information-driven stories about the world around us. Such a flood of information, of data that can be captured about certain areas of interest extends as well to the realm of the humanities, as objects of study - be they works of art visual, literary, musical, poetic or otherwise - are everywhere sealed in envelopes of metadata that allow one to aggregate, parse, and discover patterns amongst many (thousands) of like and similar works of art.

It really is an exciting time, but also a somewhat disruptive time, in the academy. Johanna Drucker, a leading thinker and longtime practitioner of what one can call digital humanities, can pen a thoughtful and thought-provoking essay for how humanists might approach dealing with data in ways fundamentally different from that of practitioners of the hard sciences (her coining of the term "capta" is particularly effective). But to judge from the reaction in the readers comments section of a recent article in the Chronicle of Higher Education in which Drucker's efforts at UCLA and the efforts of others elsewhere to lay the groundwork for a digitally-inflected humanities at the undergraduate level, it would seem that there is a good-deal of skepticism and resistance from within the Academy by fellow Humanities scholars. None of this is a surprise and it is instructive that parallels are quite apparent between the Academy and Journalism. 

It probably will be the case that that which now is greeted with skepticism will in less than a decade be an accepted facet of practice within fields of study (while not comprising the whole of). Art history undergrads will have a course (perhaps optional, perhaps mandatory) of study in statistics (all the better to understand and to generate information visualizations). Faculty, grads and undergrads may well work in collaborative teams investigating certain big questions, questions that may even cut across disciplines. All of this will just be the growth of and change in our field(s). Exciting? Yes. Uncertain? Of course. Coming? Well, there's something happening here, what it is will one day be clear (again, apologies to Buffalo Springfield).

Quint

So, will A(ugmented) R(eality) be that pervasive, that successful?
Tuesday, February 25, 2014

This past week we kicked off in the Collaboratory a Spring workshop series on Augmented Reality in museums and galleries. Think a combined salon of ideas and maker sessions and you get some idea of what John Shipman, Director of the Art Gallery here on campus, and I hope will come of these five workshops. As it was our first meeting the opportunity loomed large to introduce ourselves and jump a bit into the AR for museums/galleries question(s) with a host of links to projects and perspectives that already exist in the museum world. It was a wonderful, stimulating, fun conversation and I, for one, look forward to the workshops in coming weeks. The focus of these workshops, I think, will shift from such a heavy emphasis on the discussion of the idea of AR in museums and galleries to striking a balance with making AR objects and trackables for actual use. We have a deadline, as many of us hope to include an AR project in the Maryland Day (April 27) visitor experience to the Meme exhibition in the Art Gallery! If you are interested in joining these workshops, please do not hesitate to be in touch with John Shipman ( jshipman@umd.edu ). If you'd like to be a part of the group's listserv but cannot make the workshops (see flyer here) please direct an e-mail to me (Quint) ( quint@umd.edu ).

So, you may be wondering about the image above. That is an avatar of James May, a host on BBC's Top Gear, serving as a sort of surrogate for the curator/docent in London's Science Museum. The pattern on which he stands is a trigger for his appearance on one's iPad, iPhone or android device/tablet while at the same time one can see in the picture frame the object about which he holds forth (the content being also triggered by the pattern). Here it is just a couple of feet, which gives a sense of his tiny scale. We had a good discussion about the obviousness of this type of AR intervention and wondered about just how gimmicky it is. My guess is that what works to engage an audience in the Science Museum of London, a recognizable media personality holding forth knowledgeably about artifacts of industry and with a bit of humour, will not work in an art museum or a gallery. That question is one of the many I hope this workshop will both raise and consider.

Oh, and while probably it will not end up as my contribution to the Art Gallery's What's in a Meme? exhibition, I have a goal for our next workshop to have an unimpressed Mckayla Maroney popping up all over the place!

(An actual meme, but wait until you see how I augment the online Sistine Chapel with this concept!)

Quint

Day 2 - THATCamp liveblogging 2
Tuesday, February 11, 2014

Jessica Landau, and Melissa Seifert // “Learning to See Systems: addressing the role of vision in new technologies”

The question: "how can we make visible the values and epistemologies embedded in technological systems?"

Seven faculty (and graduate students about 10) from Univ. Illinois constructing a two-year program.

Four seminars and weekly labs - systems and how they works. Actively interrogating all of the systems of art history(ies). Construct their own visualizations in labs.

Collaboration, collaboration, collaboration!!!

Scalar published papers working through systems one of the first tasks by students within the program.

Melissa Seifert discussing her work (on epehemeral protest signs): digital space (search) augmenting the physical space capture of experience. Ethical issues to this scholarly activity?

Jesssica Landau working on hunting visual culture - notes that Scalar allows one to link to anything that is online. Makes visible an unknown artist. Can also make invisible those collections/works NOT online. gives example of Winald Reiss, whose comparatively well-known, but underrepresented online. Such lack of access influences the course of scholarly inquiry and papers written? Strange turn, no?

Quint

 

Day 2 - THATCamp liveblogging 9
Tuesday, February 11, 2014

John Resig - Computer Vision Techniques and Image Analysis for Art History Research

Super-psyched for this one!

Going over what works now for computer vision (numbers, face recognition (Open Biometrics), flat, textured objects - which requires a lot of training)

Computer Vision

Two types - unsupervised vs. supervised

unsupervised (comparing an entire image, categorizing an image)

supervised (requires labeling)

unsupervised requires little to no prepping of data, just get started, limited results

supervised training (lot of preplanning: thousands of images, metadata rich); crowd sourcing may come into play; results can be quite interesting, quite compelling, but again, thousands of images to train

Okay, tools:

imgSeek (OpenSource) compares entire image, finds similar images (not exact); does not find parts, color sensitive (B and W may not work so well); does not do the needle in the haystack search (particular details) well

TinEye MatchEngine compares portions of images; finds exact matches; finds images inside other images; color insensitive (much more flexible; looks for features of images; can find flops)

search by image (say with a snapped photo) (citing the Anonymous Italian Art project at the Frick Photo Archive) TinEye Match Engine

starts to break down in terms of matching a portion to the whole at about 30% of overall image

Image Categorization

Deep Neural networks (think Google)

requires minimal categorization; little user input

Erstaz (still in beta) (requires a lot of training) seems to be able to start to make stylistic analogues and categorize (about 60% accurate in John's test runs to this point) (more images for each artist would help)

OpenCV and CCV are gold standards for general computer vision (these require coding, whereas others do not require coding)  (object detection - not so good with 3D objects; long runs of computer processing time; tens of thousands of images)

Learn more about Computer Vision http://cs.brown.edu/courses/csci1430/

http://ejohn.org/research

what projects might be good for this?

motifs within illuminated manuscripts (Alex Brey - tried and abandoned)?

what sort of training for the computer/software necessary? tens of thousands stressed again as a standard

John is starting to work with Met on mis- or un-attributed works to group/cluster by stylistic charcteristics; project at NYTimes where they correlated known and unknowns to replace bad with better

english broadside ballad archives - NEH-funded computer vision project at UC Santa Barbara

biodiversity heritage library - illustrations from their digitized library - another NEH funded project

attribution; correcting possible cataloging/digitization issues; linked open data merging (Ukiyo-e project cinches through comparison of like images the problem of linked open data; it establishes the link and them from there works backwards, bypassing the problem of data and different institutional standards)

Is there a visual equivalent of topic modelling?

Figuring out what is possible and communicating that to and from art historians is a key thing, underscored by call for applications to NEH for workshops/projects to explore this area

In response to a question about the need for learning code, John makes the excellent point that he hopes that people will get enough training in programming/code so that they can understand the appropriate tasks for a computer and be able to dialogue with potential collaborators on the programming side of things.

Quint

 

 

 

 

Pages