Google surprises from time to time with a new fascinating results tool. His latest effort: the development of Artificial Intelligence and the possibility of looking at a photograph and knowing where it was taken thanks to PlaNet.
Identifying a specific place in an image is easy sometimes: Eiffel Tower, Statue of Liberty… but what about less common and less known landscapes or streets that we photograph?
The PlaNet program uses artificial intelligence network technology to analyze images and make appropriate estimates about where in the world the image is taken. In fact, Google researchers continue to improve the program with more than 91 million images from Street View along with the rest of the data associated with the different locations.
How does PlaNet work?
The program designed by the computer Tobias Weyand it has a simple operation: PlaNet, when you see a photograph, analyze and match pixel by pixel with the large database, it has more than 91 million geopositioned photographs. Then, by comparing image patterns, you will be able to identify the location where the photo was taken with great precision.
A striking fact, echoed by MIT’s online magazine, Technology Review, is that, during its first tests, when the database only had 2.3 million images, the artificial intelligence developed by Google was able to identify the country where the photo was taken in 28.4% of cases; and the continent, in 48%. At the moment, the estimates and hopes of the developers of Sillicon Valley are very encouraging, due to the support of the database that does not stop growing day by day.
In fact, getting what PlaNet is doing is very complicated. And to realize this difficulty, you only need to take a game to Geoguessr and see how a human is unable to guess where the images are taken.
Of course, PlaNet will not be able to repair architectural details, vehicles, signs… as do people who look at different locations through Street View.
At the moment, there are still no plans for this Artificial Intelligence system to become part of the range of Google products, but if one day it arrives we will no longer be able to deceive our followers on social networks about the exact situation of the photographs that we upload to the network.