FACE RECOGNITION USING ARTIFICIAL NEURAL NETWORK Russian patent published in 2020 - IPC G06K9/62 G06N3/02 G06T1/40 

FIELD: data processing.

SUBSTANCE: invention relates to face recognition. Such technical result is achieved due to that automatic face recognition is performed by convolutional neural network containing groups of layers, wherein the first, second and third groups comprise a convolution layer, a feature grouping layer based on the function of extracting the maximum and an activation layer of the elements using the parametric linear function, fourth group of layers comprises a convolution layer and an activation layer of elements by a parametric linear function, the fifth group of layers comprises a fully connected layer which outputs a result comprising a feature vector which characterizes the image, availability of a database containing said feature vector, together with a plurality of other feature vectors characterizing other images containing persons, a search mechanism in said database and a comparison device which determines a measure of similarity between a subject feature vector and at least one reference feature vector and compares said measure of similarity with a similarity criterion, wherein matching with the similarity criterion indicates that the person is recognized.

EFFECT: technical result is providing automatic face recognition.

14 cl, 8 dwg