Resumen
In this paper, the task is to obtain informative data using video analytic approaches based on computer vision and deep neural networks, as well as their further analytical evaluation in order to automate the process of calculating traffic in Moscow?s densely populated areas. An approach to solving this problem is proposed, allowing to calculate in the selected location two different traffic: human and automobile. A data set is needed for the neural network learning procedure. The accuracy of the algorithm is evaluated by measuring the quality measure of the detector. The paper deals with the task of determining (identifying) each individual object (person and vehicle) and further analyzing its movement. It lies in the fact that for a given video file is required to determine the number of people, vehicles moving in one of several predetermined directions. In the future, these data are analyzed (correlated with time intervals, help to identify "popular destinations", "hot zones" (zones of increased interest)), and on the basis of them build graphs and heat maps. This task relates to the field of detection of objects in the image