Most cameras and smartphones include the option to add metadata to the images, which include information about the location. Thus, one can determine exactly the location where a photo was taken. But Google has found a way to determine the location even without the meta-data.
The new project of the company is called “PlaNet”, and is able to look at an image, to analyze elements in the decor, and then to realize where that picture was taken.
The system uses a database of 126 million geotagged photos organized into 26,000 grids from sites such as Google, Street View, to identify striking similarities. PlaNet can identify writing on signs, building styles, vegetation, and many other things. It can even guess the location of indoor locations, but the idea it is not fully implemented and still needs a lot of polish.
PlaNet has a success rate of only 10% to guess where the pictures were made, when it comes to cities, probably because most of them tend to look the same, especially those on the same continent.
What’s surprising is that the system can identify indoor pictures, both through analysis of what you might find on the walls, as well as by referencing other images in your album, to determine if they are in the same location.
The system is not currently available to the public, is still in testing. And before that happens, you can be sure that some government agencies already have access to it.