Google recently trying to make a proprietary search engine as smarter as possible . For example, software companies have started to teach search engine to determine the place where the picture was taken using a smartphone or camera. For this neural network, called PlaNet was organized. Its database has been downloaded more than 90 million images from the locations in which they were photographed. Apparently, the database will be constantly updated and used as the primary resource for the new search engine option.
At the same time during a search query the coordinates photo PlaNet will need to compare the image with the image from the database at the pixel level. Thus, if the maximum number of matches the system will display the intended location of the picture. Currently, intensive work is underway to implement a new technology, so the accuracy is poor. During the first tests PlaNet managed correctly guess the only continent in the 48% and the country - in 28.4% of cases.
Related Products :
|