Maxim Mirolyubov, "Center 2M": "Authority" of video analytics becomes a driver of market growth

Maxim Mirolyubov: These are completely different concepts. Computer vision is a technology for obtaining information from video images and pattern recognition. Machine or technical vision is a specific method of applying computer vision, for example, in industrial tasks - in enterprises where high-precision control is required.

Video analytics is a technology for analyzing the array of images that we obtain using computer vision. Based on such an analysis, it is possible to draw some conclusions, make decisions or delegate them to higher-level systems.

Maxim Mirolyubov: This question cannot be answered unequivocally. There are tasks that allow the simultaneous use of both technologies. It is difficult to say which method will give the best result. Video analytics on the server is quite popular, since most video cameras do not have enough powerful chips. Devices can solve one or more relatively simple problems (for example, detect the intersection of a line, movement). But the cameras still lack the resources to process complex scenes "on board."

If the video stream is processed at, there are servers virtually no limits on computing power. Video stream from one camera can handle different neural network models, solving several problems of different levels and complexity at the same time. That is, they can detect a person, monitor the presence of personal protective equipment or get into a danger zone.

Perhaps the most optimal approach is to create complex solutions using both types of video analytics.

Maxim Mirolyubov: For companies like Center 2M, that is, working with large Russian enterprises, the on-premise model is now more in demand. But for small companies, where it is important for business to get a service in a short time without large investments in infrastructure, the cloud solution is very effective. That is why "Center 2M" in this regard is ready to offer different service models for different customers.

Maxim Mirolyubov: One of the most popular functions is facial recognition. Now solutions based on this technology are offered by many companies, there is more or less stable demand from the business. The technology is mainly used in the field of security and access control, to count unique visitors in retail and take into account the working time of staff.

Another popular basic functionality technology is line crossing detection, intrusion into the zone, person counting, license plate recognition. Its popularity is due to the fact that this functionality is often on board many intelligent video cameras and does not require much resources.

More complex functions include, for example, monitoring the availability/absence of personal protective equipment on staff members. The functionality of our solution CenterVision in this area allows you to detect several elements at once: helmet, safety glasses, gloves, overalls. We can also combine different detectors to execute a certain scenario (for example, monitor the presence of gloves only when working with equipment in a certain area).

Maxim Mirolyubov: Of course, in the first place is the smart city system. For our state, security is a priority, so systems with recognition of people, faces and non-standard behavior have great prospects. Many of these solutions are already used, take at least a biometric facial recognition system in the Moscow metro.

Video analytics for industry has serious prospects. Machine vision cameras are used to detect defects, control the shape or geometric dimensions of products. As part of this approach, Center 2M has developed a module for reading data from markings for accounting and monitoring process progress in production.

There are good prospects in agriculture, for example, in animal husbandry. An example is the joint project "Center 2M," IBM and one of the largest agricultural holdings. As part of the project, we try to track the signs of disease in animals using video analytics. Detection of diseases at an early stage will significantly reduce the company's losses, lead to an increase in the productivity of the entire agricultural complex.

to trade Retail also actively uses various types of video analytics: to control the layout of goods, cash centers, etc. Many retail chains already use such pilots. We assume that in the near future retailers will begin to fully implement video analytics systems in their stores.

Maxim Mirolyubov: The volume of the video analytics market in Russia in 2019 amounted to about 19 billion rubles. The growth forecast reached 20%, but how the situation with stock markets and the economic situation in the world will affect this forecast remains to be seen.

So far, one of the main prerequisites for growth, in our opinion, is digital transformation projects, which now exist in many Russian enterprises. In such projects, video analytics is one of the key tools for monitoring, optimizing and digitizing business processes.

A good sign is a gradual decrease in cost and an increase in the availability of computing capacity. This is due to the fact that some video analytics scenarios require the processing of video streams on-line and the output of the result in almost real time. If many cameras are involved - 50, 100 or more, then the cost of resources increases by a multiple. Previously, this made customers think a lot, but now the situation is changing, because they not only see a real prospect for the development of their businesses from the introduction of technology, but also understand how this can help break away from competitors. That is why we are now seeing an increasing number of requests in the company, including those not related to security. The "authority" of video analytics becomes a growth driver.

The more successful and large-scale implementations, the more market growth we will see in the future.

Maxim Mirolyubov: One of the main factors, in our opinion, is a possible increase in cost. Tasks that must be solved in real time require large computing capacity, and therefore other investments from the customer.

As for protection personal data and systems built on identifications biometric signs, according to 152 Federal Law "On Personal Data" video, by default is not a personal medium, data because it is not used for identity identification. The video just contains an image of a certain person. You can store such data without restrictions. This area certainly requires a deeper legal settlement, and the state has not yet said the last word on this issue.

Maxim Mirolyubov: There are a number of such companies - they, like "Center 2M," are distinguished by the ability to form working solutions for any, even the most difficult tasks. Other companies are engaged in video analysis only as part of the construction of video surveillance systems. They could also be attributed to serious players, but such companies are used to focusing on security-related tasks. Creating diverse solutions is not an advantage.

Maxim Mirolyubov: It is difficult to compare the maturity of the world and Russian market, since the segmentation of solutions and specific tasks solved using these products is quite large. We believe that Russia is at the beginning of the journey. But our experience shows a steady growing interest in technology from customers. We constantly explore new cases and observe a positive trend and interest in technology not just as a new, incomprehensible tool, but already in terms of pilot implementations and deployment of machine vision systems throughout the enterprise. It can't help but please.

Maxim Mirolyubov: Most of the requests received by our company are usually associated with five main scenarios. This is the detection and counting of people, recognition of the presence/absence of personal protective equipment or other visible objects on a person, control of the location of objects in a given area, recognition of faces, symbols. We have models and modules for each of these scenarios. But this is more of a basic functionality, a kind of cubes, from which we collect solutions for the specific task of the customer.

The control of behavior and type of activity is very interested in our law enforcement agencies, which ensure safety in transport, in places where mass events are held.

Video analytics for monitoring traffic is growing in popularity. On the agenda is the development of a system of full-fledged classification of vehicles. Such a system will allow you to determine the number, model, color - and collect everything in a single database of incidents or violations.

Maxim Mirolyubov: This will be possible only if the demand comes from government agencies. For example, a company like Rostelecom is unlikely to give anyone access to its cameras. However, there is a reason in this approach. You just need to remember that the requirements for the characteristics of the cameras and their placement differ depending on the task that this camera solves. That is, it will be very difficult to recognize faces from a camera installed to cover a large territory and count people on it.

Maxim Mirolyubov: Our decisions are based on what is now called "artificial intelligence" - convoluted neural networks, mathematical models trained to recognize certain images on pre-prepared material. The better this material and the larger it is, the better the result of the model.

Maxim Mirolyubov: We work with IBM. Our cooperation began from the moment we came to the market. It turned out to be a really effective interaction: IBM has created a powerful tool for creating neural network models that our analysts use to solve problems in this area.

But our collaboration is not limited to the use of technology. We are implementing many joint projects. For example, now the Center 2M, together with IBM, is developing a fundamentally new product - a modular video analytics platform that will solve almost any problem related to video recognition. At the same time, the platform initially includes the possibility of integration with IBM solutions, in particular with the BI (Business Intelligence) system.

Maxim Mirolyubov: It would be great if the customer company had a comprehensive description of its business processes. Comprehensive analytics provide a quality solution. In addition, this approach avoids unnecessary waste and calculates in advance whether the video analyst can solve the task or not.

In order for the client to work effectively, a team must participate in the project. Ideally, this is the so-called "iron triangle": leader, expert and analyst.

Manager - has a complete vision of the situation and is responsible for communication between the team and developers. It sets tasks for project participants, monitors their implementation and determines priorities. If the team does not have a leader or his role is "blurred" between many employees, then a well-known situation will turn out.

The expert is an employee who understands in detail the problem process. The task of the expert is to explain what and how it works and what abnormal situations can await the customer and the developer.

A business process analyst is someone who can describe a customer's business process in detail, calculate costs and correlate them with the possible benefits derived from implementing the technology.

Is it possible to assign all one hero instead of three participants in the "iron triangle"? In general, it is possible, but the experience of implementing the projects "Center 2M" shows that it is better not to do this. All three roles require different competencies. A project manager is usually a manager.

Expert - a narrow specialist with experience. Analyst - expert with special competencies and knowledge necessary for preliminary analysis.

And finally, the main condition for success. You need not be afraid to try, look for and test new solutions. We are always ready to show how the technology works, advise and suggest the individual solution required by a particular customer.