In this week’s installment of welcoming our new computer overlords, Google now can identify exactly where you were—anywhere in the world—when you took a scenic photo. PlaNet, the adorable name for our eventual downfall as a human race, is the tech giant’s newest deep-learning program that is able to decipher where an image was taken without geo tags. The key to this insane advancement is feeding PlaNet all of the photos from Google’s massive library of over 90 million Google Maps images.
Tobias Weyand and his software engineer team behind PlaNet used neural network technology to ensure that the program can recognize images from places without landmarks—like a dusty dirt road or an untouched field. The Google program could identify the location of an image all the way down to the street only 3.6% of the time, but its accuracy grew as the net was cast wider like at city level (10.1%), country level (28.4%), and continent level (48%). While these numbers seem small, just remember that almost 4% of the time a computer can tell exactly where you were on Earth when you took a photo with no prior knowledge.
PlaNet is still in its development stage, but it boggles the mind to think about where this technology can take us. Movie buffs could really pinpoint their favorite releases’ filming locations down to the icy patch in which Leonardo DiCaprio ate that bison liver. And Carmen Sandiego? Your thieving days are numbered.
(via The Independent)
