FIELD: image processing means.
SUBSTANCE: present invention relates to adjustment of three-dimensional face model to two-dimensional face image. In the disclosed method of processing a two-dimensional image: obtaining a two-dimensional image; processing a two-dimensional image by a trained convolution neural network; creating a three-dimensional face model and superimposing the created three-dimensional face model on the face area in the two-dimensional image. Disclosed inventions enable to embed a three-dimensional transformable avatar into a two-dimensional image in real time on a mobile computing device of a user.
EFFECT: providing more stable and correct adjustment of a three-dimensional face model to a two-dimensional image, in which angles of rotation and/or inclination of the human head relative to the chamber are significant.
7 cl, 5 dwg
