CapsNet comparative performance evaluation for image classification

TL;DR

This study compares CapsNet's performance with traditional classifiers across various image datasets, finding that while CapsNet requires high computational resources, its potential for future improvements remains promising.

Contribution

The paper provides a comparative analysis of CapsNet against established classifiers, highlighting current limitations and future potential in image classification tasks.

Findings

CapsNet requires significant computational resources.

CapsNet's accuracy is below that of Fisher-faces, LeNet, and ResNet.

Further research with enhanced architectures may improve CapsNet performance.

Abstract

Image classification has become one of the main tasks in the field of computer vision technologies. In this context, a recent algorithm called CapsNet that implements an approach based on activity vectors and dynamic routing between capsules may overcome some of the limitations of the current state of the art artificial neural networks (ANN) classifiers, such as convolutional neural networks (CNN). In this paper, we evaluated the performance of the CapsNet algorithm in comparison with three well-known classifiers (Fisher-faces, LeNet, and ResNet). We tested the classification accuracy on four datasets with a different number of instances and classes, including images of faces, traffic signs, and everyday objects. The evaluation results show that even for simple architectures, training the CapsNet algorithm requires significant computational resources and its classification performance falls below the average accuracy values of the other three classifiers. However, we argue that CapsNet seems to be a promising new technique for image classification, and further experiments using more robust computation resources and re-fined CapsNet architectures may produce better outcomes.