Structured neural network for recognition of shapes

A short description of a neural network model is given in this article. The model is intended for the recognition of object shapes. The formulation of the problem assumes that all areas with uniform texture and brightness are segmented in a gray-scale object image and therefore their boundaries are extracted as well. The extracted object contours are input to the model. A well-known example of such a problem is the problem of handwritten character recognition. A number of different methods have been proposed for solution of this problem. In this paper an attempt is made to solve it by means of a structured neural network with learning