The Cotswolds as seen through Google Dream
- Credit: Archant
Every image run through Google’s Deep Dream algorithm emerges as a psychedelic vision of Dali-esque surrealism. Can you tell where these iconic Cotswold locations are after they’ve been given the treatment?
Image recognition software is advancing in leaps and bounds and is now found on everything from your favourite social media platform to your digital camera. The engineers at Google have been tinkering away in an attempt to improve their software, which has inadvertently spawned a viral trend. The Deep Dream algorithm.
Essentially, Google want to teach a program to be able to analyse, distinguish and classify objects in a fundamentally human (or animal) way - being able to tell if a spade is a spade... or actually a trowel, or a rake etc. This is done through teaching the program using reference images in much the same way we learn to understand the world around us.
The Deep Dream algorithm is what happens when the program looks at an image and is asked to find objects within it. It then amplifies these features over and over until the objects it’s looking for are more visible. So even with an image of white noise, it analyses the arrangement of dots, compares this with what it has learnt from millions of reference images, and starts to ‘connect the dots’, so to speak. Doing this multiple times gives you a bizarre amalgamation of the original image and various ‘imagined’ objects. Because the program is biased towards identifying certain things, just as we are naturally biased towards identifying human faces, the images regurgitated by it often feature dogs, birds, eyes, cars and so on.
Now everybody has been applying the algorithm to their own images - spawning psychedelic, hallucinatory and hilarious artwork.
We’ve run pictures of a number of well-known places in the Cotswolds through the algorithm. Can you identify the locations?
Put your own pics through the process using dreamdeeply.com
Did you correctly guess the locations from the pictures? Have a look at the original images here.