Can you outwit Google’s new photo-identifying neural network?

Using all these images, Weyand has taught a powerful neural network to zero in on the grid location itself without any human intervention.

And to further evaluate the system’s complex power (which we can call the fifth part of the development), the team tested it with 10 well-traveled humans.

Photos are broken down to the pixel-level, and Google’s systems then try to cross-reference these with its huge image library to see if it can match it. In order to test the system, Google scientists loaded PlaNet with 2.3 million geotagged Flickr photos and asked it to work out where they were taken.

To kick off the project, the team divided the world into grids consisting of 26,000 squares of varying sizes.

In a trial run using 2.3 million geotagged images, PlaNet determined the country of origin with 28.4 per cent accuracy and the continent of origin in 48 per cent of cases.

The test numbers are significantly better than humans performing the same test.

PlaNet outperforms humans and it can even recognize the location of indoor photographs and specific things such as pets or food.

The team behind the AI is made up of a pair of software engineers at Google, Tobias Weyand and James Philbin, and developer Ilya Kostrikov. Then they created a database for PlaNet that contained 126 million geolocated photos pulled from the Internet, Technology Review reported.

Geotagging images may be a chore, but it’s frustrating where you come across an image online and you can’t tell where it was taken. Google PlaNet is not 100 percent flawless, yet, but it is continually improving and learning from its past mistakes at determining where photos have been taken, which might prove to be helpful not only in helping individuals remember where certain photos were taken, but also in aiding the police find criminals, and the government to locate terrorists.

Images often contain informative cues such as landmarks, weather patterns, vegetation, road markings, and architectural details, which in combination may allow one to determine an approximate location and occasionally an exact location.

“PlaNet won 28 of the 50 rounds with a median localization error of 1131.7 km, while the median human localization error was 2320.75 km”, according to the paper.

Even more fantastic is the fact that PlaNet only takes up 377 MB of space, which can snugly fit into smartphones. And that’s not at all too surprising – the point of A.I. isn’t to fundamentally mimic the human brain in all ways, but to surpass human limitations in a few specific ways to accomplish much more hard tasks. The network has been loaded with over 90 million photos that have been tagged with locations.