Who installs the cameras and who uses them?

Smart Safe City

The Kipod system, on the basis of which the Republican Public Safety Monitoring System (RSMOB) operates, is a typical Smart City platform. What are the challenges for smart safe cities? For example, there is a list of wanted persons: missing persons, criminals, suspects, people under house arrest. The task of this system is to ensure the processing and search of these persons in real time, and, if necessary, then later to ensure the search for video in the archive from these hundreds of thousands of cameras. And it is desirable that it be faster than the whole life.

Yes, the system can search for previously identified people. She does not need to know who's name, but if there is a photograph, then you can find out if a person has appeared in those places where the cameras are. But Kipod doesn't know how to identify everyone passing under the camera (technically).

For example, they say: let's load the entire database of passports into the controlled list, and since he knows how to identify people from this list, then let him do it. However, one must understand that the passport database is always larger than the number of citizens in the country, it can store canceled, expired documents. If we assume that the population is 8-9 million, then the base will be about 30. If you do not go into super-complex matters, the idea is as follows: each picture you have is, in fact, a descriptor (a mathematical description of a face for a neural network) and you have a task comparing descriptors, which is a little more complicated than comparing the same strings or numbers. We have a comparison speed of approximately 200 thousand descriptors per second. As soon as there are more of them, time increases linearly.

Even if we assume that not the entire database of passports is loaded, Minsk passports were taken from it, some other narrowing criteria were introduced, and there are not 9 million, but a million of them, all the same, we get that each person will be processed for 5 seconds. The system will eventually "choke".

Using Kipod

Why do we know that our system was not used to recognize protesters' faces? I will not dive into everyday logic about how many cameras there are and where they are. Protest actions take place in different parts of the city and, at least, it is impossible to outweigh the cameras in the hope that you have guessed where this time everything will happen.

How are citizens identified in practice? In various sources: social networks, telegram channels, the media - there are photos from the actions. These photographs are used by a number of software tools to search the passport database. If necessary, additional data are requested from cellular operators: where the subscriber was and at what time. All this is described in detail in protocols and regulations, right down to which computer was connected and which IP address was assigned when analyzing the photos.

Who installs the cameras and who uses them?

Nobody collects data for the sake of data. It does not make sense. Some practical problem needs to be solved, and these systems (for smart cities) really help to solve them. Statistics show that most crimes are of a domestic or near-domestic nature. People often do something based on the belief that they will not be found. When there is a system that ensures that any speeding you make will have consequences, people start to violate less. As a result, many lives have been saved.

Nobody sticks their fingers into the socket in the hope that they usually do not shock them, but if they hit, is it okay? Everyone knows for sure: stick your fingers in - it will be bad. When people understand that a violation of the speed limit or any other rules will always, in 100% of cases, be recorded, people stop violating.

All provisions on the system are available in the public domain, not classified as secret, and everyone who is very interested in it finds the data, where it is written who is the user of the system and on what conditions. Users are not established by us, but by law; even we are not users of the system, we are only responsible for its performance. We have a commercial relationship with subscribers who own cameras: subway, railway and so on. Is our system used by state bodies of Belarus? Exactly yes, because government agencies are engaged in the tasks assigned to the Republican system for monitoring public safety voice gateways cisco.

Once again: those cameras that are connected to the system belong and are available to the subscriber for his own use, and for use within the framework of the RSMOB, by system users (government agencies, within the framework of the tasks they solve, described in the regulation on the system). We don't buy, install or use cameras.

Obviously, we are not the only company in the world that implements systems for safe cities. There are N companies in each country that are engaged in similar tasks.

But few of them are interested in the tasks of mass identification of citizens - it is very expensive and has no practical sense. Just take a Google photo, upload photos from your passport database, and then take the photo you are looking for and put it in your album. Google will suggest matches it finds. And it is difficult to modify Kipod for mass identification, this task was never before us and, accordingly, was not implemented.