SOLVING THE TASK OF FACE RECOGNITION IN CASES OF INSUFFICIENT TRAINING SET.

This paper describes methods that are aimed to solve face recognition tasks with an insufficient training set. These methods include: creation of a 3D model of a head that is based on a basic training set - three images of faces (profile, half turn, full face), placing and analyzing control points on a model, calculating distances between points (the ones not used in the creation of a model), which is followed by face recognition. The created 3D model allows acquiring additional images of faces (at different angles), which significantly increases the results of recognition of unknown faces, as compared to only using the basic training set. The proposed methods were tested on various images of faces. The results have shown that these recognition methods can be used in cases, when the initial information about the shape of the face is insufficient, for example, in forensics. [ABSTRACT FROM AUTHOR]