Can you guess where the above picture was taken? Google can.
Yes, Google is close to applying finishing touches to a revolutionary new tool, dubbed the Google PlaNet, that will tell you where a photo was taken by simply looking at its contents, according to a report published in PC Mag. The new tool will not need any extra data to figure out the location where the image was taken, provided that the image itself is not blurry or poorly shot, the report adds.
Making use of ‘superhuman’ artificial intelligence, the team developing Google PlaNet trained it to identify locations where photos were taken based on visual cues. According to Technology Review, the database for the Google PlaNet contains a staggering 126 million geolocated photos pulled from the Internet.
Of course, this is not the first time that Google has attempted to figure out where an image might have been shot just by its contents. For humans, it would be a terribly difficult thing to do, as was proved when Google tested the probability via the website GeoGuesser, which tasks people with figuring out exactly what PlaNet has to figure out: where are you?
Although not yet completely perfected, Google PlaNet is proving to be considerably better than humans at figuring out the geolocation of a photograph, according to Tobias Weyand, a Google specialist whose team built PlaNet.
“We think PlaNet has an advantage over humans because it has seen many more places than any human can ever visit and has learned subtle cues of different scenes that are even hard for a well-traveled human to distinguish.”
The median human localization error has been approximated to be 1442 miles (2320.75 km), whereas Google PlaNet brings that median error considerably down to 703.2 miles (1131.7 km). “This experiment shows that PlaNet reaches superhuman performance at the task of geolocating Street View scenes,” Weyand added.
TechSpot reports that to measure Google PlaNet’s accuracy, the team fed it 2.3 million geotagged images from Flickr and other photo-sharing websites, and the results were greatly encouraging for the future of geolocation. Researchers are particularly proud of the fact that artificial intelligence has been able to outsmart human intelligence, even when it is trained solely on image pixels and geolocations.
“One would expect that these cues, especially street signs, together with world knowledge and common sense, should give humans an unfair advantage over PlaNet, which was trained solely on image pixels and geolocations. Yet PlaNet was able to outperform humans by a considerable margin.”
Another reason that Google PlaNet has performed much better than expected is because researchers trained the machine to do ‘sequence location’ — that is, exploiting the way photos are often taken in sequences and so the system uses entire photo albums — in cases where the visual cues in photographs was not substantial, according to ZDNet.
Google PlaNet has been constructed in a way that uses as little memory as possible — only 377 MB to be precise — meaning it can easily be downloaded on a smartphone.
It is still not clear exactly when Google plans to launch the tool for the general public, but considering there is still scope for further improvement in the technology, it would be safe to assume that Google PlaNet is still some way off from becoming an indispensable tool on our smartphones.
Having said that, it would be interesting to see what other advances Google can make now that they have successfully tested the superhuman A.I.
Google PlaNet, at least so far, has been a roaring success.
By the way, have you still been able to guess where the featured image in this article was photographed?
[Photo of Austria, Valley, Europe, Alps/Pixabay]