SYSTEM FOR RECOVERY OF ROCK SAMPLE THREE-DIMENSIONAL STRUCTURE Russian patent published in 2020 - IPC G06T17/00 G06N3/10 

FIELD: physics.

SUBSTANCE: invention relates to computer engineering and can be used to construct three-dimensional digital models of structure of rock samples from two-dimensional images of flat sections of rock. System comprises a data storage and processing device, a computing module, an encoder neural network, a generator neural network and a discriminator neural network. Encoder neural network is configured to receive, convert 2D images of flat slices of the rock into a vector of parameters and send the converted slices to a generator neural network which is configured to recover the synthetic images of the three-dimensional structure of the rock sample based on the parameter vectors, and the discriminator neuron network is configured to compare synthetic images of the three-dimensional structure of the rock sample with real images of the three-dimensional structure of the rock sample stored in the training sample.

EFFECT: technical result is higher accuracy of restoring three-dimensional structure.

14 cl, 2 dwg