Image Super-Resolution Using VDSR-ResNeXt and SRCGAN

TL;DR

This paper introduces two novel super-resolution methods, VDSR-ResNeXt and SRCGAN, combining deep convolutional networks and conditional GANs to improve image quality and speed, validated on standard benchmarks.

Contribution

The paper presents two new super-resolution techniques that integrate ResNeXt architecture with VDSR and a class-conditioned GAN approach, advancing the state-of-the-art.

Findings

Both methods outperform previous models on benchmark datasets.

VDSR-ResNeXt achieves higher accuracy with efficient computation.

SRCGAN improves perceptual quality through class conditioning.

Abstract

Over the past decade, many Super Resolution techniques have been developed using deep learning. Among those, generative adversarial networks (GAN) and very deep convolutional networks (VDSR) have shown promising results in terms of HR image quality and computational speed. In this paper, we propose two approaches based on these two algorithms: VDSR-ResNeXt, which is a deep multi-branch convolutional network inspired by VDSR and ResNeXt; and SRCGAN, which is a conditional GAN that explicitly passes class labels as input to the GAN. The two methods were implemented on common SR benchmark datasets for both quantitative and qualitative assessment.