The paper offers an overview of modern text-to-image graphical neural networks and text-to-image image retrieval. The paper discusses a number of problems arising in the use of text-to-image networks and possible methods for their solution. One of the actual tasks related to the study of graphical neural network technologies becomes the study of neural network images and identification of images obtained with the help of neural networks among other graphical content. We propose approaches to solving problems aimed at solving the problem of verification of media materials and developing algorithms to detect artificial (neural network) origin of photo and video materials. The rapid development of neural network technologies in this area can have a significant impact on society, the market of professions and the media, which makes the task of identifying neural network images among other graphic content particularly relevant.

Machine learning, computer vision and pattern recognition, neural network, computer graphics, Text-to-image

Bondareva Nikole Alexandrovna,  ,  orcid.org/0000-0002-7586-903X,  KIAM RAS