Analysis of the Basic Image Generation Methods by Neural Networks
Timur Fatkhulin, Yuri Leokhin, Maxim Mentus, A. A. Kulikova, Rafif Alshawi
Abstract
The article compares the main existing neural network methods that allow you to get images by entering a text request in an application or online service. The purpose of the work is to analyze modern neural network methods for generating images. The relevance of this work is due to the high demand for the use of images generated by neural networks in many areas of human life, such as advertising services, goods, business processes, animation and design. Achievement of this goal is ensured by solving the following tasks: consideration of on-line services based on different methods of image generation, determination of their advantages and disadvantages, identification of general patterns of image generation for each method. The basic principles of image generation by each of the considered neural network methods are presented. The methodological basis of the article is a descriptive method, methods of theoretical analysis, as well as a generalization method.