CLASSIFICATION OF SMALL DATA SETS OF IMAGES WITH TRANSFER LEARNING IN CONVOLUTIONAL NEURAL NETWORKS

  • Biserka Petrovska Military Academy "General Mihailo Apostolski" UGD
  • Igor Stojanovic
  • Tatjana Atanasova Pacemska

Abstract

Nowadays the rise of the artificial intelligence is with high speed. Even we are far away from the moment when machines are going to make decisions instead of human beings, the development in the field of neural networks is remarkable. They are in a big expansion in a new millennium. Their application is wide: they are used in processing images, video, speech, audio and text. One special kind of neural networks, convolutional neural networks have been a point of interest of researches in the last decade. These networks have been widely applied to a variety of pattern recognition problems, such as computer vision. Convolutional neural networks were trained on millions of images and it is difficult to outperform the accuracies that have been achieved. Sometimes we have a small set of images to be classified and in those situations there is no success to train the network from a scratch. This article exploits the technique of transfer learning for classifying the images of small datasets. It transfers the knowledge of the pre-trained convolutional neural network and use it for the classification of those data sets. Fine-tuning of the network is done through optimization of hyper parameters, in order to maximize the classification accuracy. At the end, the directions have been proposed for the selection of the hyper parameters and of the pre-existing network which can be suitable for transfer learning.

Published
Aug 7, 2018
How to Cite
PETROVSKA, Biserka; STOJANOVIC, Igor; PACEMSKA, Tatjana Atanasova. CLASSIFICATION OF SMALL DATA SETS OF IMAGES WITH TRANSFER LEARNING IN CONVOLUTIONAL NEURAL NETWORKS. Balkan Journal of Applied Mathematics and Informatics, [S.l.], v. 1, n. 1, p. 17-24, aug. 2018. Available at: <http://js.ugd.edu.mk/index.php/bjami/article/view/1948>. Date accessed: 22 oct. 2018.
Section
Articles