Google has succeeded in recognizing images

Scientists of Google created a program that knows how to most effectively recognize objects in the photographs.




The program as described this picture: "Two pizza on a plate».

Descriptions of images, programming, not much inferior descriptions composed by people.

In addition to facilitating opportunities to find photos, the program also opens up new opportunities to help blind people in the perception of pictures.

Artificial intelligence program created by Google, uses two types of neural networks - for recognition and to describe their natural language. A neural network is a computer model, partly repeating unit of the human brain. These systems consist of interconnected neurons, which can receive information from multiple sources and capable of learning.

Google created a neural system is the brainchild of four scientists - Oriola Vinyalsa, Alexander Toshev, Samy Benji and Dumitru Erhan.

"It is better to see once than hear a hundred times," - says they blog .

"But sometimes words are more useful, so we need to learn how to translate words into images in automatic mode" - the scientists write.

Two years ago, researchers have created a Google program to recognize the images that "show" of 10 million frames of video clips stored on video sharing YouTube. Three days later, the program learned to allocate the array of images of cats.

Source: geektimes.ru/post/241726/