Coverspace

This is a quick little project that came about at ThatCamp New England in October 2013. For a more technical slant, see the project page for coverspace on GitHub.

Humanists are often quite good at dealing with large, unstructured collections of texts; we’re less good, often at looking at images.

Lev Manovich and the Institute for Software Studies developed ImagePlot, which places large collections of texts by their image characteristics. (How dark they are, how much red, etc). There’s a lot to like about this approach–at the very least, it starts to raise all sorts of interesting questions about what sorts of facets of images computers are good at detecting, and what they want to see. It makes movies, but there’s no reason not to build the images created this way directly into the infrastructure of the open web.

At ThatCamp New England in October 2013, several of us–including Peter Leonard and Lauren Tilton–were talking about whether that approach could be made more accessible. Web browsers seemed like the way to do it. Browsers are fine-tuned, complex, and omnipresent tools for handling images; thanks to Mike Bostock’s D3 library , it’s easy to position lots of them in the screen. That makes it easy to share visualizations of a set with others; to manipulate it on your own machine; and to recreate it anywhere else.

It’s extensible. At a first pass, we simply grab the colorspace configurations for an image across some of the most common elements–but the basic model could be extended to entropy.

This example works with OWI photos, but you can run it on your own as well. Just fill a directory with jpg files and run “make”, and the program will take care of the rest. (Well, really the program will die for lack of dependencies; but you’ll have to install those.)

Someone would probably have to grant-fund this to make it into a more general purpose tool. It could be a pretty interesting Omeka plugin, for example, to try exploring different facets of a library; if some of the image analytics were done in javascript instead of imagemagick/python as they currently are, you could even create a bookmarklet that would instantly pull all image files from any web space and let you dance them around.

All this is contingent on how useful it is to begin with. That seems like an open question. But it’s fun to zip images around the screen, at least.

There’s plenty left to be done–most importantly, filtering out topics by metadata.