Google has created a unique algorithm describing the images, which can be downloaded by developers using open source software library – TensorFlow. After an appropriate training algorithm can recognize objects shown in pictures with an accuracy as high as 93.9 percent. This is a significant improvement over the technology two years ago, when artificial intelligence was able to correctly classify 89.6 percent of photos.
It is better described photos could be used in many ways – for example, assist historians, to the blind, and other companies working on artificial intelligence. Such an algorithm is therefore a very useful solution. The algorithm of 2014 years using the method of image classification called Inception V1. The algorithm of this year, while the use of a more advanced version of this method – Inception V3. Thanks to artificial intelligence can not only recognize what is in the pictures, but also understand what dependencies exist between the visible elements.