Group Equivariant Networks Using Morphological Operators

With the increase of interest upon rotation invariance and equivariance for Convolutional Neural Network (CNN), a fair amount of papers have been published on the subject and the literature keeps increasing. This paper aims to fill the lack of morphological approaches on the matter. We propose a set of group equivariant layers using morphological operators, several model configurations are tested and compared with a convolutional equivalent network. The results show that the proposed morphological networks are capable of classifying rotated images even when trained only with upright samples.