Instruments and Systems: Monitoring, Control, and Diagnostics Annotation << Back Neural Network Identification of the Mark of Blanks SPC-1 JSC «OEMK» D.A. POLESHCHENKO, I.S. ZORIN This article is devoted to the development of neural network deep learning algorithm for decimal character recognition. It is shown that together with the developed criteria for localization of the recognized object, this approach allows you to accurately determine its location in the image. High accuracy of positioning on an object increases the efficiency of its identification.The algorithm is based on the developed convolutional neural network of deep learning. Deep convolutional networks have been very successful in teaching tasks that provide unprecedented computer vision performance. The article analyzes the effectiveness of digit recognition convolutional neural network deep learning, as well as analysis of the impact of neural network architecture on the effectiveness of character recognition. It is determined that convolutional networks with sufficient accuracy recognize the numbers in the image, despite the factors distorting the original image. Training and testing of neural networks was carried out on the basis of SVHN images and this was sufficient for effective recognition without creating your own database of objects. Keywords: Neural network, deep learning, pattern recognition, convolutional neural network, segmentation, binarization. DOI: 10.25791/pribor.06.2020.1184 Pp. 32-39.