A Survey on Capsule Networks

Abstract

In today’s digital era, machine intelligence is equally as important as human intelligence. The emergence of deep learning techniques makes machine intelligence tasks easier and better. Deep convolutional neural networks are prominent in tasks of object detection, image classification, object segmentation, and so on. Recently, Hinton and his team introduced a new architecture called capsule networks. Capsule networks replace the neurons with capsules and overcome spatial and rotational invariance limitations of convolutional neural networks. This paper defines the introduction and working of capsule networks with related work in the field of capsule network.