Let's get away from flashy posters with "unique super-analytics" and "all-seeing technologies". We try to understand and fix what exactly is planned to control with the help of video analytics.

The essence of the step is that for robotic control using video images, it is necessary to have a clearly defined "norm" of the controlled business process, or in other words, its regulations. This allows to set and formalize control parameters, signs of events, etc.

Here it is important that the regulations should not be written on paper, but real. It really has to define the actions of the company's employees at the controlled objects.

If there are no such regulations, you have to develop and implement them. You can't do anything without it!

The result is a significant reduction in financial costs.

The need for such a step is due to the market strategy of many companies offering the implementation of neural network video analytics. They rigidly link innovation with the installation of new video surveillance system elements. Moreover, it is not uncommon that with this approach, ordinary cameras suddenly become highly intelligent and get a price tag that is three times higher than the cost of analogs that are not branded with artificial intelligence.

To avoid such significant overpayments, you need to assess the capabilities of your own equipment. You just need to determine the type of cameras - IP or analog, in the existing video surveillance infrastructure. Depending on the result there will be two options:

• (a) Analog type. Upgrading to IP cameras is inevitable;

• (b) Digital type. You should open the archive and establish whether there are recordings of the processes to be monitored (see "step one"). If there are such recordings with sufficient resolution (you can see the necessary details), then congratulations! Your video surveillance system allows you to implement neural network analytics without a significant upgrade of the final equipment (cameras).

This step contains a tactical technique that allows you to evaluate the prospects of cooperation with a potential contractor. To implement the technique, it is necessary to prepare video fragments of the processes that are supposed to be sent to the neurorobot for cross-checking (see "Step Three", point b). These fragments will make the negotiation substantive and help to bring the positions closer together more quickly. At the same time, it is important that they should be in a "native", not compressed format (resolution).

The purpose of this step is to result in reports with the required information in a suitable and auditable format.

This step is also related to the market practice of a number of service providers. For example, a group of companies offers services in the field of retail. The services include facial recognition for traffic recording, control of repeated visits and the facial database is limited to 10 employees of the firm. Why only 10? Because an increase in the facial base by at least 2 times would require an order of magnitude more complex algorithms.

As a result, the reports from such contractors contain a photo of each entrance (can be solved by the regular features of the budget camera), but the graphs of repeat visits are abstract. In fact, instead of personalizing traffic and analytics on repeat visits, such companies give customers graphs with information that cannot be verified.

This paragraph is about the traditional choice: software purchase or subscription fee? It's up to you, but keep in mind that subscriptions have been winning in the software world for a long time now. In addition, experience shows that the choice in favor of subscription fees makes the contractor much more loyal in fixing bugs and support.