Data Augmentation for Foreign Material Discrimination using Deep Learning

Abstract

The authors try to discriminate small foreign materials using a convolution neural network (CNN). In this paper, the authors verify the effects of data augmentation for discriminating small foreign materials. The authors experimented for three target foreign materials, woodchips, pieces of plastic and cymothoidaes. The authors measured accuracy of discrimination for foreign materials without data augmentation and with it. The results of the experiment proved that data augmentation contributed to improvement in accuracy of discrimination for small foreign materials.