MIT Technology Review today detailed Google’s new neural network for determining the location of images without geotags.
This project led by Tobias Weyand, a computer vision specialist at Google, has potential to revolutionize your photo library. The machine, in its current state, is able to significantly outperform a human’s ability for recognizing the location of an image. It even has a procedure for determining the location of indoor images, where there are absolutely no clues as to where the image could’ve been taken. The human eye wouldn’t be able to determine a location like that, but Google’s new neural network can.
Granted, it’s not perfect yet. This AI–called PlaNet–in a trial run with 2.3 million images was able to determine the country of origin 28.4 percent of the time. Similarly, it was able to determine the continent of origin 48 percent of the time. While that’s not high on the accuracy front, it’s significantly more accurate than a human trying to determine these values.
Here’s the full breakdown of how it works, per the MIT Tech Review:
“The team created a database of geolocated images from the Web and used the location data to determine the grid square in which each image was taken. This data set is huge, consisting of 126 million images along with their accompanying Exif location data. Weyand and co used 91 million of these images to teach a powerful neural network to work out the grid location using only the image itself. Their idea is to input an image into this neural net and get as the output a particular grid location or a set of likely candidates.”
PlaNet is quite impressive in its current state, but the possibilities down the road are endless. Imagine PlaNet becoming much more accurate than the current percentage rates, and then having something like that embedded in Google Photos. And yes, it’s quite possible to be implemented into a smartphone to work with something like Google Photos.
“Our model uses only 377 MB, which even fits into the memory of a smartphone,” Weyand said. It’s still quite far off, but once again, the possibilities for PlaNet in the future could eventually revolutionize how we use our photo galleries.
For the full rundown on PlaNet, be sure to hit the source link below.
source: MIT Technology Review
via: The Next Web
Come comment on this article: Google unveils new neural network, can determine location of images without geotags
from Android News, Rumours, and Updates http://ift.tt/1R7rDOZ
SOURCE- talkandroid
from Creptico http://ift.tt/24t5Pr3
via IFTTT
EmoticonEmoticon