Effect Analysis of Loss Function for Image Super-Resolution Based on Improved ESRGAN

Miao Pan

2023

Abstract

The discipline of picture Super-Resolution (SR) has experienced an exceptional advancement with the development of deep learning. The Generative Adversarial Network (GAN) has emerged as the most popular deep learning technique for super-resolution. In order to create a super-resolution model, this study presents Enhanced Super-Resolution GAN (ESRGAN). Loss types covered include Content Loss, Adversarial Loss, and Total Variation Loss (TV Loss), with a focus on how the ESRGAN model affects picture super-resolution when applying various loss functions. Then, many ESRGANs with various loss functions were compared and assessed using common image evaluation indicators including Peak Signal-to-Noise Ratio (PSNR), Structure Similarity Index Measure (SSIM), and Learned Perceptual Image Patch Similarity (LPIPS). It is discovered via the comparison of trials that changing the Wasserstein GAN (WGAN) adversarial loss from the ESRGAN adversarial loss may significantly increase the stability of GAN network training. By modifying the loss function, the enhanced ESRGAN suggested in this research may successfully enhance the super-resolution impact of pictures.

Paper Citation

in Harvard Style

Pan M. (2023). Effect Analysis of Loss Function for Image Super-Resolution Based on Improved ESRGAN. In Proceedings of the 1st International Conference on Data Analysis and Machine Learning - Volume 1: DAML; ISBN 978-989-758-705-4, SciTePress, pages 149-153. DOI: 10.5220/0012804500003885

in Bibtex Style

@conference{daml23,
author={Miao Pan},
title={Effect Analysis of Loss Function for Image Super-Resolution Based on Improved ESRGAN},
booktitle={Proceedings of the 1st International Conference on Data Analysis and Machine Learning - Volume 1: DAML},
year={2023},
pages={149-153},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012804500003885},
isbn={978-989-758-705-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Data Analysis and Machine Learning - Volume 1: DAML
TI - Effect Analysis of Loss Function for Image Super-Resolution Based on Improved ESRGAN
SN - 978-989-758-705-4
AU - Pan M.
PY - 2023
SP - 149
EP - 153
DO - 10.5220/0012804500003885
PB - SciTePress