FIELD: physics.
SUBSTANCE: invention relates to a text field identification mechanism. Method includes obtaining one or more hypotheses for the type of the field of the first text field present on the image of the document and creating a three-dimensional matrix of features which represents part of the image containing the first field. Text field identification mechanism provides a three-dimensional feature matrix as input data for the trained machine learning model and obtains output data from the trained machine learning model, wherein the output data contain a quality estimate of one or more hypotheses.
EFFECT: technical result consists in expansion of arsenal of means for identification of text fields.
20 cl, 8 dwg
