We've been playing around with TinEye, the 'reverse image search' (registration required). As of this evening, the site claims to have crawled 1,013,140,121 images, assembling a giant database that can be used for near-instant comparison. From the FAQ: 'TinEye finds exact and altered copies of the images that you submit, including those that have been cropped, colour adjusted, resized, heavily edited or slightly rotated. TinEye does not commonly return similar matches, and it cannot recognize the contents of any image. This means that TinEye cannot find different images with the same people or things in them.'

The site does a good job of pulling up a set of differently sized, coloured and scaled versions of the same painting. Maurice de Vlaminck's Landscape with Red Trees (1906) gives the above set of thumbnails a ripple of difference - admittedly mostly very slight - but noticeable in terms of hue and crop. But what about paintings by the same artist? Or different versions of the same landscape? (Paul Cezanne painting Mont St Victoire, for example). Or even different views painted using the exact same combination of colours? Imagine if it could be set to find works by the same artist working in a similar way? TinEye could not only help research artistic movements, it could uncover potentially hidden works. It could create new movements.


Above, a TinEyed selection of thumbnails of one Cezanne painting. Below, several thumbnail images of paintings of the same view, all by Cezanne.



But what about brands? Could TinEye be trained to identify a Nike trainer, regardless of model, a BMW, or even a building by Frank Gehry? Repetition breeds familiarity in the world of branding, but the idea that an object's inherent brand values might be digitally quantifiable opens up huge cans of worms for product designers. All things seem possible. Imagine the launch of truly recognition engine, a new business tool that is seen as the litmus test for brand recognition. Simply upload the design, adjust the sliders, and you can whether or not your design has _enough_ BMW in it through it's ability to 'attract' and be associated with existing products.



If you run a search, pick 'closest match last' to see how images - usually stock or press shots - are clipped, chopped and pasted. These tiny deviations from the original are examples of the emerging digital patina, the inadvertent introduction of imperfections through the encroachment of jpg degradation, crops and colour recalibration. The inability of digital art to replicate itself precisely is referenced in recent work by Thomas Ruff (sometimes v.nsfw). Ironically, the very tool that reveals this hitherto visual richness in digital design might ultimately lead to the push-button blandification of the material world.

Labels: art, design, technology

posted by things at 00:13 / 3 comments