Retrievr: Visual search of Flickr

It’s been noted on other blogs throughout the week, but I wanted to mention Retrievr, Yet Another Clever Flickr Interface (YACFI?), as it relates to the problem of burrowing through large sets of images. A Flash interface lets you make simple sketches, the color palettes of which Retrievr then attempts to match with images from Flickr (details from the developer here). See also comments from ResearchBuzz (“[E]ven if it doesn’t work very well it makes an awesome toy and you should go play with it right now”).
We’ve experimented with visual search for image retrieval, adding technology from Idée to the organizer component of Photoshop Elements. Elements lets you choose as many as four source photos in the browser window, then automatically find other photos similar in appearance. Elements 4 also introduces some new face-tagging features created by Adobe’s Advanced Technology Group.
Incidentally, if you’re wondering why we sometimes introduce new technology in Elements before bringing it to Photoshop proper, think of it as similar to, say, the Foo Fighters dropping in to play an out-of-the-way club. It’s a chance to introduce promising, useful work without immediately subjecting it to the full glare of hardcore pro workflows.
[Thanks for the Retrievr link to Mark Kawano, formerly the UI designer for Camera Raw, now at frog design until we can steal him back. ;-)]

0 thoughts on “Retrievr: Visual search of Flickr

  1. That is absolutely fantastic! I’m mezmorized how the heck it can do that. I drew a simple profile of a man and it works….i thought it had to do w/ the flickr api but it’s not — it’s python driven madness!

Leave a Reply

Your email address will not be published. Required fields are marked *