In Russia, a neural network with a “human” vision was created

Scientists from the Institute of Mathematical Problems of Biology of the Russian Academy of Sciences have created a neural network that manages its “look” and looks for objects on the perceived picture in much the same way as the organs of sight and the human brain do, according to an article published in the journal Neural Networks.

“The developed model offers a simple and unexpected explanation for a very complex cognitive process of searching and recognizing objects in the picture perceived by our eyes,” says Yakov Kazanovich from the Institute of Mathematical Problems of Biology of the Russian Academy of Sciences in Pushchino. According to him, the neural network created by his team should help neurophysiologists to understand how real human eyes work.

Over the past decade, hundreds of programmers and dozens of large IT companies have created countless machine vision systems that can recognize and classify different objects in a perceived picture. These data, modern robots, search engines and drones can use for a variety of purposes – for example, to avoid obstacles or find a client when delivering a parcel.

Despite the tremendous progress in this field, scientists still do not actually know anything about how the vision of humans and animals works and how we manage to automatically classify and recognize even those objects that we have never seen before.

Therefore, as Kazanovich says, many features of human consciousness, perception of reality and vision are still a mystery for neurophysiologists and psychologists. For example, scientists argue long ago about why a person very easily finds “contrasting” objects in a huge variety of other structures unlike him, but at the same time finds it difficult to search for several figures hidden in a small number of similar objects.

Kazanovich and his colleague Roman Borisyuk made a big step towards solving this problem by creating an artificial intelligence system that behaves in the same way as a person when solving these problems.

Its main feature, as the scientists say, is that it consists of a set of relatively independent from each other structures, the so-called “ensembles”, neurons in which produce special fluctuations. One of these structures becomes a kind of “conductor”, which manages the work of other “ensembles” and distributes tasks to them, and other ensembles are in fact objects that “see” the neural network in the picture.

“Ensembles” constantly compete with each other for their influence on the “conductor” and on the work of the entire network as a whole. The way this competition is going on, as shown by experiments and calculations by Kazanovich, almost perfectly reflects the principle of the work of human vision and it is similar to the “slip” of our view on the picture when searching for objects of varying degrees of “contrast”.

This model, scientists hope, will help neurophysiologists not only find similar structures in the human brain and monkeys, but also understand how they work, which will bring us closer to creating “natural” machine vision systems.