Google now automatically recognizes the place in the photo! Google introduced an advanced system that allows it to find out where the selected photo was taken. Now you no longer need geotagging or geotagging a photo so Google can locate it. He does this with the PlaNet computer neural network, which searches through more than 90 million photos with known coordinates, finds the best pair and determines the location based on that. It is a learning process, which means that Google is continuously learning and adding to its knowledge (read: filling the database).
Google now he recognizes himself the place in the photo, but it is not necessary for this geotagging. This enables him PlaNet neural network, which is pre-equipped with more than 90 million photos with a location badge, on the basis of which he understands where the selected photo was taken, and he completes his knowledge of the place with each new photo.
READ MORE: What happens if you zoom in on photos on Google Maps too much?
The first attempt at a computer neural network was made by Google on 2.3 million randomly selected photos, which proved to be extremely reliable. She correctly determined the continent in as much as 48 percent, and the country in 28.4. She did a little worse with streets, as she guessed them correctly in only 3.6 percent of cases, and cities in 10.1 percent. Harder than it seems, huh?
If you don't think the data is impressive, it should be added that the database was large in the test 377 MB, so the experts agree that the system hides in itself huge potential. Otherwise, if you think you're better than Google, try it out Geoguessr, let's see you, heroes.
More information:
arxiv.org
