Automatic recognition of objects in the image is considered to be quite difficult even for computers with special algorithms and support of the person. Recognition absolutely all objects using a computer at the moment is almost an impossible task, but in a short time, this process can become a reality thanks to the efforts of Google on the development of new recognition system that will be able to identify most of the objects, even if they are partly covered by other objects.
The essence of the system is a neural network that can determine the search criteria without the use of large computing power. Such a system is capable of much more in-depth scan of the image and the definition of several objects on the basis of assumptions. At the moment, this technology can easily identify most of the things in the living room, including a "flying cat". How Google is going to use this product - is still unknown.
Developing recognition technology research department engaged Google Research Lab. On the official website you can find more details about the system.