FACE RECOGNITION USING IMAGES BINARY CLASSIFICATION METHODS.
Stochastic separation theorem applied to face recognition task using Fisher’s linear discriminant is presented in this paper. The theorem says that in spaces with high dimension there is a high point dividing probability. There is an idea about using feature vectors as a points in a high dimension spaces. As a method to take feature vectors convolutional neural networks are used. In this paper Inception ResNet v1 neural network architecture is used with 128-dimenstion feature vectors. As it can be seeing, precision of this method is equal to 0.9980 and recall is equal to 0.9623.
Authors: A. M. Golubkov
Direction: Informatics and Computer Technologies
Keywords: Neural networks, binary classification, stochastic separation theorem, linear discriminant, Mahalanobis distance
View full article