Group Equivariant Convolutional Neural Networks for Color Fundus Images Super-Resolution

Juntao Jiang, Yong Liu

Keywords: Fundus images, Super-resolution, Group Equivariant Convolution Neural Networks, MSRResNet

TL;DR: Proposed a MSRResNet with Group Equivariant Convolution Neural Networks for fundus images super-resolution

Abstract: High-resolution (HR) color fundus images can provide finer details and help with a more accurate diagnosis, while deep learning-based super-resolution methods usually require a large amount of data. This paper integrated Group Equivariant Convolution Neural Networks (G-CNNs) in Modified SRResNet (MSRResNet) for fundus images super-resolution. Experiments showed that it leads to some improvement compared with regular CNNs.