Tourists often have to remember where it is one or the other picture of the trip was made. Soon, this will help them developed by Google PlaNet system, which is capable of independent learning and can recognize an image at the pixel level. Comparing the data obtained with the already available information, the program determines the location where the picture was taken.
According to Lonely Planet, the system is still in its infancy. But at least, it has in some cases surpasses human capabilities in determining the place where the picture is no obvious landmarks.
According to the developers of the project, people can use these tips as those or other natural objects, weather, vegetation, road signs and architectural features to remember about the location. The new technology uses a similar algorithm. Photo of a typical beach can be made in a variety of locations around the world. If in this case the picture is not specific to a particular place objects, people can pay attention to details such as street signs, while PlaNet uses for this pixel.
In this program, the entire world is a network with thousands of geographic locations. These cells are correlated with photos containing Geotags – information about the place where the picture was taken. It uses information about the millions of photos posted on the Internet. The system then identifies those or other details. In some cases, it shows a man inaccessible level of accuracy.
The advantage of artificial intelligence is the ability to analyze a virtually unlimited amount of information, whereas a person based on his limited experience. Also, the system is able to remember minor details that are difficult to see even a careful user.
As proof of the effectiveness of its development, the creators of PlaNet conducted tests. The program competed with ten experienced travelers in the online quiz Geoguessr, in which you need to guess the picture in the photo space. The program has prevailed in 28 of 50 rounds.