Similarity Maps

Similarity maps illustrate the complex relationships that exist between a single image and the rest of the Glocal image pool. The image at the centre of each map is the seed. Images from the pool that are most compositionally similar to that seed image are connected to it by a series of lines.


Images that are similar to this second ‘generation’, then, are positioned in branches that radiate away from the seed. These particular maps show images that are within 4 degrees of similarity to the seed. These structures are unique - each of the 11,000+ images in the pool will construct a different map when used as the seed. Also, these similarity maps will change over time, as more images are added to the pool and more similarities are identified. In this way, these maps can be thought of as temporal fingerprints of each image and their context within the pool.

These similarity maps were constructed using a custom-written software tool built in ActionScript 3.0. The similarity engine is driven by LibPuzzle, a library for identifying similar images, written by Frank Denis. 

The Glocal Similarity Map engine will be released for public use in the coming weeks, along with full source code. For more information about, please contact us:

Glocal Similarity Map Engine

Glocal Similarity Map Engine Close-up

Glocal Similarity Map Engine Detail