Підвищення ефективності формування сцен доповненої реальності за допомогою нейронних мереж

Abstract

On the way to the fourth wave of industrial technological progress, visualization and virtualization tools have received a wide range of applications and integration into the multi-industry space. The technology of creating additional visual images is currently used in the medical field, the field of education, the industrial and industrial field, the field of advertising and trade, in the field of modeling and design, in the scientific field, the cultural and entertainment field, etc. The potential of using visualization tools is inexhaustible, because the integration of additional information in the form of graphic objects helps to increase the perception of the data flow of reality and develops analytical capabilities for users of augmented reality technology. Modern means of creating scenes of augmented reality and additional visual images have increased requirements for the consumption of computing power, as they require dynamic adaptive interaction with streams of real data, which actually leads to the formation of extremely complex algorithms and corresponding hardware-analog and software-digital solutions. Optimizing and improving the efficiency of the augmented reality scene creation technology is a scientific problem that needs to be solved, including within the scope of the current research. According to the bibliographic search and analysis of modern trends and profile developments, the potential possibility of using neural network tools to create additional visual objects in augmented reality scenes has been established. Neural networks have a high adaptive capacity for learning and an adequate reaction to external conditions of functioning. Therefore, neural network tools are surprisingly suitable for integration into technological solutions for the functioning of augmented reality technology. Known topological solutions for arranging and organizing the functioning of neural networks, which can be applied to solve a certain scientific problem of optimizing the consumption of computing power and increasing the efficiency of creating augmented reality scenes, have a number of limitations to their application, which prompts the further search for adaptive solutions. A promising solution is the formation of combined-hybrid technologies for constructing the topology of neural networks. Thus, the relevance of the research is outlined, the scientific issues are formed, and the vector of scientific research to solve the identified issues is proposed.