Інформаційна технологія нейромережевої ідентифікації людини по фото при трансформації зовнішності

Abstract

The information technology of identification of the person on a photo at transformation of appearance which is based on the convolutional neural network of ResNet is offered. The software implementation of human photo identification was performed using face image descriptors from the Dlib library, the Skimage library for downloading photos from files and the SciPy library for calculating the Euclidean distance between descriptors. The developed program has a higher reliability of human identification by photo (95%) than the best of similar programs BioID (91%), i.e. the reliability of human identification by photo is improved by 4%.