Structure Analysis and Performance Comparison of Image Generation Methods Based on Generative Adversarial Networks

ZiRui Wang

2024

Abstract

In recent years, Generative Adversarial networks (GAN) have made remarkable progress in image generation. This paper reviews various GAN-based image generation methods, including CycleGAN, Pix2pix, and StarGAN models, focusing on their performance on different tasks and data sets. The advantages and limitations of each model are discussed by comparing structural similarity (SSIM) and peak signal-to-noise ratio (PSNR). Comprehensive experimental data analysis results show that different GAN models behave differently in specific application scenarios, CycleGAN performed well on image diversity tasks, Pix2pix has an advantage in high-fidelity scenes, while StarGAN shows excellent performance in face image generation. In this paper, the characteristics and application scope of each model are summarized, and the development direction of image generation technology in the future prospects, including model fusion, high-resolution image generation, multimodal fusion, and so on. This study is designed to act as a guide for researchers and practitioners in the field of image generation and to encourage the expansion and implementation of generative adversarial network technology.

Paper Citation

in Harvard Style

Wang Z. (2024). Structure Analysis and Performance Comparison of Image Generation Methods Based on Generative Adversarial Networks. In Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML; ISBN 978-989-758-754-2, SciTePress, pages 45-50. DOI: 10.5220/0013487100004619

in Bibtex Style

@conference{daml24,
author={ZiRui Wang},
title={Structure Analysis and Performance Comparison of Image Generation Methods Based on Generative Adversarial Networks},
booktitle={Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML},
year={2024},
pages={45-50},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013487100004619},
isbn={978-989-758-754-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML
TI - Structure Analysis and Performance Comparison of Image Generation Methods Based on Generative Adversarial Networks
SN - 978-989-758-754-2
AU - Wang Z.
PY - 2024
SP - 45
EP - 50
DO - 10.5220/0013487100004619
PB - SciTePress