|By Kramer Phillips | 4 years ago|
Google, one of the most forward-thinking companies in the world, has done it again with their new concept PlaNet. It does not need GPS or Geotags to depict the location of the photo, not even data from your camera. All it needs is the content of the picture itself. The concept is still far from completion, but the results so far have been incredible.
It is astonishingly difficult to know the location of a photo as most land masses look identical over a wider view. Google’s first experiment with their device was through one of their websites to see, in relation to human beings, how rigorous their program was.”In total, 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,” said Tobias Weyand, a Google analytic.”This small-scale experiment shows that PlaNet reaches extraordinary performance at the task of geolocating Street View scenes.”
The program is surprisingly data efficient, needing only 377 megabytes to run. The program was created by dividing the world into 26,000 grids. It then loaded 91 million images to the program and denoted where each image is located on the world network. The image-grid database would then give the location of the picture acting as the data center. They then added another 34 million images from all over the world to increase accuracy.
Its first experiment was to use the program with Geotagged images on Flickr. From the 2.3 pictures, PlaNet got an accuracy level of 3.6 percent on street level. It also got a 10 percent accuracy on the city, 28 percent of the country and 48 percent of the continent.