# DeepCaps: Going Deeper with Capsule Networks

## TL;DR

DeepCaps introduces a deeper capsule network architecture with a novel 3D convolution routing, achieving superior performance and fewer parameters on key datasets, while enhancing interpretability through a class-independent decoder.

## Contribution

It presents DeepCaps, a deep capsule network with a novel 3D convolution routing algorithm, surpassing state-of-the-art results and reducing parameters significantly.

## Key findings

- Outperforms previous capsule networks on CIFAR10, SVHN, and Fashion MNIST.
- Reduces model parameters by 68%.
- Enables control of image attributes via the decoder.

## Abstract

Capsule Network is a promising concept in deep learning, yet its true potential is not fully realized thus far, providing sub-par performance on several key benchmark datasets with complex data. Drawing intuition from the success achieved by Convolutional Neural Networks (CNNs) by going deeper, we introduce DeepCaps1, a deep capsule network architecture which uses a novel 3D convolution based dynamic routing algorithm. With DeepCaps, we surpass the state-of-the-art results in the capsule network domain on CIFAR10, SVHN and Fashion MNIST, while achieving a 68% reduction in the number of parameters. Further, we propose a class-independent decoder network, which strengthens the use of reconstruction loss as a regularization term. This leads to an interesting property of the decoder, which allows us to identify and control the physical attributes of the images represented by the instantiation parameters.