Is the deconvolution layer the same as a convolutional layer?

TL;DR

This paper clarifies the relationship between deconvolution and sub-pixel convolution layers, demonstrating that low-resolution space convolutions offer better representation power at the same computational cost.

Contribution

It provides a detailed analysis of different convolutional layers, especially the efficient sub-pixel convolutional layer, and compares their effectiveness in low-resolution versus high-resolution spaces.

Findings

Low-resolution space convolutions have more representation power.

Efficient sub-pixel convolutional layer is distinct from standard deconvolution.

Networks with LR space convolutions outperform high-resolution upsampling at fixed complexity.

Abstract

In this note, we want to focus on aspects related to two questions most people asked us at CVPR about the network we presented. Firstly, What is the relationship between our proposed layer and the deconvolution layer? And secondly, why are convolutions in low-resolution (LR) space a better choice? These are key questions we tried to answer in the paper, but we were not able to go into as much depth and clarity as we would have liked in the space allowance. To better answer these questions in this note, we first discuss the relationships between the deconvolution layer in the forms of the transposed convolution layer, the sub-pixel convolutional layer and our efficient sub-pixel convolutional layer. We will refer to our efficient sub-pixel convolutional layer as a convolutional layer in LR space to distinguish it from the common sub-pixel convolutional layer. We will then show that for a fixed computational budget and complexity, a network with convolutions exclusively in LR space has more representation power at the same speed than a network that first upsamples the input in high resolution space.