APSIPA Transactions on Signal and Information Processing > Vol 8 > Issue 1

Checkerboard artifacts free convolutional neural networks

Yusuke Sugawara, Tokyo Metropolitan University, Japan, Sayaka Shiota, Tokyo Metropolitan University, Japan, Hitoshi Kiya, Tokyo Metropolitan University, Japan, kiya@tmu.ac.jp
 
Suggested Citation
Yusuke Sugawara, Sayaka Shiota and Hitoshi Kiya (2019), "Checkerboard artifacts free convolutional neural networks", APSIPA Transactions on Signal and Information Processing: Vol. 8: No. 1, e9. http://dx.doi.org/10.1017/ATSIP.2019.2

Publication Date: 19 Feb 2019
© 2019 Yusuke Sugawara, Sayaka Shiota and Hitoshi Kiya
 
Subjects
 
Keywords
Convolutional neural networksUpsampling layerCheckerboard artifacts
 

Share

Open Access

This is published under the terms of the Creative Commons Attribution licence.

Downloaded: 6447 times

In this article:
I. INTRODUCTION 
II. PREPARATION 
III. PROPOSED METHOD 
IV. EXPERIMENTS AND RESULTS 
V. CONCLUSION 

Abstract

It is well-known that a number of convolutional neural networks (CNNs) generate checkerboard artifacts in both of two processes: forward-propagation of upsampling layers and backpropagation of convolutional layers. A condition for avoiding the artifacts is proposed in this paper. So far, these artifacts have been studied mainly for linear multirate systems, but the conventional condition for avoiding them cannot be applied to CNNs due to the non-linearity of CNNs. We extend the avoidance condition for CNNs and apply the proposed structure to typical CNNs to confirm whether the novel structure is effective. Experimental results demonstrate that the proposed structure can perfectly avoid generating checkerboard artifacts while keeping the excellent properties that CNNs have.

DOI:10.1017/ATSIP.2019.2