The work on the Painter Goblin is nearly done for now. The automation of his work is almost ironed out with regular tweets happening every night.
I just wanted to tie off this series of blogs with some open thoughts.
Uncompressed, Van Gogh’s Starry Night on Wikidata is approximately 1.8GB, 30,000 x 20,000 x 3 (RGB). This breaches the Twitter image upload limits. There’s a temporary fix in place to handle this based on file size, where the image’s resolution is reduced by the percentage difference required to meet Twitter limits. Twitter however seems to rely on both file size and resolution – and so I may swap in the correction algorithm that manages the file size by resolution instead.
I’d really like to incorporate photography into this work but like drawing where I discovered there may be too much detail to create a nice effect, it may be the same in some photographs.
– it seems there are also limited public domain resources available to Wikidata – National Library of Wales seems to have a lot of drawings for example. Photos also seem limited to a handful of institutions.
If I can filter them well enough, and if I can find a decent haul, then mid 20th Century photographs onward would be great to play around with. Sid and Nancy, Nirvana, or Jimi Hendrix… imagine.
Alternative Sources of Images
In previous blogs I’ve mentioned Europeana, Trove, DPLA. If I could incorporate images from other resources into this work i’d love to promote those as well, take for example this Europeana Search for the colour Indigo.
Currently the SPARQL queries I use and mentioned in part 3, are quite well suited to works of art in museums and galleries.
If collection and location are missing from a Wikidata entry – e.g. for an entry about the performer Jeff Goldblum, then the Painter Goblin can’t pick up an image and then tweet it out.
When I have the opportunity I’ll write a heuristic or new set of queries to work around this limit – this would allow arbitrary sources of images on Wikidata to be used for fun. It may be as simple as utilising SPARQL’s OPTIONAL keyword more effectively, but we’ll see.
A hot topic is the International Image Operability Framework (IIIF) http://iiif.io/. I haven’t gained enough experience with this to understand how it can help with discoverability through consistent cataloguing of images, or discoverability across a wide number of sources. What I know it can do though is help unlock, not just whole images, but regions of images! This is quite exciting when we think about other ways the Painter Goblin may want to remix some of the images that cross his path.
I think it’d be nice to add some interactivity to this work. Allowing users to send through Wikidata links, or their own images. The former would be a great way to promote Wikidata and have them engage in a way I’m learning to. The latter, could just be good fun. I’ll likely need to change my hosting to be able to make this service happen.
All that’s left after that is to enjoy the output. If you like any of the art produced, let me know. If you’ve any suggestions or designs on palettes, or even requests for images (at least until interactivity is added), please let me know. It’s nice to know when something gives back to people. And please RT,
(Painter Goblin, https://github.com/ross-spencer/painter-goblin/blob/master/README.md)