waplog

Artificial intelligence Google will determine where the photo was taken


If you are afraid of total surveillance, then this news will hardly please you. Google is developing a new neural net PlaNet that will be able to determine the place where almost any photograph. In the process of teaching artificial intelligence as hints uses geotags, but in the future he will be on the shoulder determine the place in the photograph, without such «Cribs».

For training a neural network, Google used a collection of 90 million images tagged geotagged. The results of the first phase of testing are quite promising, because the PlaNet has already learned to identify the country with the probability of 28.4 percent, and the continent – with a probability of 48%. In 3.6% of cases even artificial intelligence defines the street where the picture was taken, and in 10% of cases is a city.

The author of the project of the PlaNet – Tobias Weyland. For the realization of the idea he shared our planet for 26,000 sectors, the size of which depends on how many pictures are made in one area or another. It is so large and populous cities covered the smaller sectors, rather than desolate and deserted. Oceans, North and South pole of the Earth the program is not considered at all. During testing of the neural network, the developers gave her more than 2.3 million photos from Flickr, taking the parallel Protocol the success of the artificial intelligence.

The results, of course, far from ideal, but the network continues to learn. To make the system even more intelligent, the developers have made it a party game Geoguessr, available at this link. During the game the user to show pictures from Google Street View, and it needs to guess the approximate place on the map where these shots were taken. When PlaNet competed online against 10-year-experienced travelers, she won in 28 rounds out of 50. Pretty good result for a beginner.

The most surprising is the fact, that the entire PlaNet is only 377 MB of computer memory that makes future quite possible its use on mobile devices. 


MTCHT
ICT
TECHNOLOGICAL INNOVATIONS
POST
ABOUT US
NEWS
INTERESTING
INTERVIEW
ANALYSIS
ONLAIN LESSONS