Source: Analyzing and Annotating Images on Thinglink
Our readings this week, based on the theme of “Beyond the Static Image.” John Resign, a developer at Khan Academy and the creator of jQuery, used image similarity analysis within the Frick Art Reference Library’s photo archive, which contains more than 1 million photographs, and his findings were pretty incredible. He wrote up his experience in the article “Using Computer Vision to Increase the Research Potential of Photo Archives”, and details how, using technology from TinEye’s MatchEngine and developing his own software, he analyzed a batch of images of anonymous Italian art. The image finder could match similar images, and at that point a human evaluator could see if the images were close enough to be matches, or a painter copying another, or simply two different paintings with a very similar composition.
One solution that mage similarity analysis in some ways helps with is the messiness of differing cataloging and metadata standards, which is a challenge we have addressed more than a few times in this class alone. As Resig explains, “The results of the image similarity analysis of a photo archive are extremely exciting and could completely change how the process of cataloging images is completed. It could also make some impossible tasks, such as merging multi-million image archives, a reality.” It is worth noting that what Resig is talking about is not linked date, but more of an image-pattern recognition tool that can help match to correct metadata, or, as he describes the three areas where image analysis would have an immediate impact: analysis and error correction, digitization, and merging archival collections.
During our classroom workshop last week we played around with Thinglink, which might be my new favorite platform for creating a digital gallery that I can, to a degree, make my own and also open up to a social media platform if people want to follow me, like my posts, or tag and annotate my posts. Due to the ease of metadata imports, the elegance of a gallery-like exhibition (Omeka) or imposing more linear story in book form (Scalar) I understand the value of an Omeka or Scalar for an end-result scholarly project. For the inspiration, workshopping, and thinking-things-out phase I really like the “mood board” or Pinterest-like feel of Thinglink. In short, I like that there are plenty of different platforms out there that can hold the different phases of a research project. Thinglink definitely has a more exploratory, less-rushed feel from which curatorial choices can then be made.
For my Thinglink example below, I chose an image of a drawing by the incomparable Indian artist Nasreen Mohamedi (1937-1990), one of those favorites I wrote an essay on a while ago, then forgot about, then she popped into my head recently. Mohamedi worked primarily in ink on paper and black and white photography. Her aesthetic is characterized by a nuanced palette of the grayscale, capturing abstractions of urban architecture in her photography as well as organic forms in her striking use of line and light drawings. Interestingly, her work has been most compared to, and sometimes exhibited with, the American painter Agnes Martin. A few years ago, she was introduced to Western audiences in exhibitions aside Martin’s paintings. It looks like that immediate comparison has dwindled since I first investigated Mohamedi, as a few years ago it was the case that a Google search on Nasreen Mohamedi yielded the suggestion: “Related Searches: Agnes Martin.” Part of the point I made in my essay was that the comparison, though legitimate in a curatorial sense, was fairly ubiquitous, superficial and lacked a context of Indian art history. The fact that Martin was a painter and Mohamedi was not brings me to the trouble of looking at Mohamedi’s images digitally: It is so hard to tell the medium of Mohamedi’s works in a Google image search! It might be interesting to create my Thinglink channels based on medium. I find I’m wrong when I try to guess whether something is a graphite drawing, pen & ink, or photography, all of which the artist gained mastery in. I used a Mohamedi image to plug into Tin Eye, which found 4 exact matches, and Google image search, which gave much broader results, as we expected.
Anyway, here is the beginning of a possible project like that, with a title that includes Artist, Title, Year, and Medium, a link to the artist’s Wikipedia page, and a link to an exhibition a couple of years ago at the Tate. When I populate the gallery more, I would like to include her own writings, as her diaries have been archived.
 Resign, John, “Using Computer Vision to Increase the Research Potential of Photo Archives.”ejohn.org. Accessed February 15, 2016. http://ejohn.org/research/computer-vision-photo-archives/