Convolutional Neural Network and Capsule Neural Network: A Survey on architecture and applications
Abstract
Convolutional neural network(ConvNet/CNN) and capsule neural network(CapsNet) are integral components of deep learning applications.Nowadays, Convolutional neural network based approach is used extensively in the broad spectrum of computer vision .Using simple computations ,the complex features are studied effectively by CNN architecture.But,of late, it has been observed that CNN has a very important pitfall of not considering the spatial relationship of data.This results in poor performance in certain applications. The deep learning research paradigm acquired a major break-through ever since the advent of Capsule networks. These networks depend on modelling the hierarchical relationships in perceiving an image much similar to human brain. This paper explains the conceptual architecture and applications of both ConvNet and CapsNet.The exploration of CapsNet and the subsequent results augurs well for the research community in imitating the human brain with much better accuracy.