IMPROVING THE ACCURACY OF CIRCULATING COINS RECOGNITION BY USING A CONVOLUTIONAL NEURAL NETWORK WITH MULTIPLE OUTPUTS
DOI:
https://doi.org/10.26906/SUNZ.2021.3.069Keywords:
convolutional neural network, multi-output neural network, image recognition, machine learningAbstract
The article considers the problem of circulating coins recognition using convolutional neural networks. A traditional approach to solving the problem of image recognition, which involves the use of a regular convolutional neural network with one output, is described, the results are shown and analyzed. To improve the recognition accuracy, the architecture of a convolutional neural network with multiple outputs was used. The results obtained were compared with the results of a regular network, the reasons for the differences in the results and the advantages and disadvantages of each of the considered approaches were givenDownloads
References
Хайкин Саймон. Нейронные сети: полный курс, 2-е изд. : пер. с англ. / Саймон Хайкин. – М. : Издательский дом «Вильямс», 2006. – 1104 с.
Ямпольский Леонід Стефанович. Нейротехнології та нейрокомп'ютерні системи : підручник / Л.С. Ямпольский, О.І. Лісовиченко, В.В. Олійник. – К. : Дорадо-Друк, 2016. – 576 с.
A Survey on Multi-output Learning / [D. Xu, Y. Shi, I. Tsang та ін.]. – 2019. – 21 с.
A survey on multi-output regression / H.Borchani, G. Varando, C. Bielza, P. Larranaga. – 2015. – 27 с.
CS231n Convolutional Neural Networks for Visual Recognition [Електронний ресурс] – Режим доступу до ресурсу: https://cs231n.github.io.
Keras API reference [Електронний ресурс] – Режим доступу до ресурсу: https://keras.io/api/.
LeCun Y. Convolutional Networks for Images, Speech, and Time-Series / Y. LeCun, Y. Benigo. – 1995. – 14 с.
Module: tf | TensorFlow Core v2.6.0 [Електронний ресурс] – Режим доступу до ресурсу: https://www.tensorflow.org/api_docs/python/tf?hl=en.
