Artificial intelligence technologies have long been used in face recognition systems and people search using CCTV cameras. However, these are far from the only parameters that can be used to search. For example, a group of researchers in India trained artificial intelligence to search for the right people based on their height, gender and the clothes they wear.
This technology may seem very strange to someone, because "recognizing" people by their faces, you can get more accurate data. But it is not so. The researchers themselves give an example. Imagine that you only know certain search parameters and an approximate location. And instead of watching all the material from all cameras, you can create a request for, for example, "women in red shirts, whose height is 153 centimeters." This will narrow the search and significantly reduce the time to identify a specific person.
The system is based on a convolutional neural network (CNN). This is a subtype of neural networks based on deep machine learning technology. CNN uses in its work some features of the functioning of the visual cortex of the brain. If you try to explain it in simple terms, there are segments that react to simple signals (for example, the presence of red) and there are more complex ones - a conglomeration of simple functions (for example, all types of shirts). Many small segments can be part of several large ones (shirts, T-shirts, pants, etc. can be red). On the construction of connections between segments, the neural network can conclude about the presence of certain objects and their properties.
As for the algorithm itself, at the moment the accuracy of its work is about 60% (the neural network correctly guesses 28 people out of 41 on average). This may not seem enough, but this is only the first version of the algorithm that will be improved. As the developers themselves stated, Vladimir Kuznetsov
