Generative AI

4Pix2pix GAN for Image-to-Image Translation: A Comparative Study with Diverse Datasets

Lakshmipriya B.^1*, Pankaj Kumar Singh² and Jayalakshmy S.³

¹School of Engineering, Jawaharlal Nehru University, New Delhi, India

²School of Computer Science and Engineering, Vellore Institute of Technology, Chennai Campus, Chennai, India

³Department of Electronics and Communication Engineering, IFET College of Engineering, Villuppuram, India

Abstract

Image-to-image translation (I2IT) is a complex task involving the transformation of images from one domain to another while retaining their essential content. With the onset of generative adversarial networks (GANs), image translation has attracted interest recently as it can synthetically generate images belonging to a new domain or scene from the images available in another domain. I2IT essentially learns the mapping of patterns from images in one domain to another domain and finds a wider application ranging from satellite images to medical image enhancement and autonomous driving simulations. Pix2pix GAN is a groundbreaking I2IT architecture that has completely changed how realistic images are produced from input data. This chapter examines the effectiveness of pix2pix GAN in producing high-quality output images from a variety of datasets, including urban planning, medical imaging, and underwater image enhancement. Through an analysis of pix2pix GAN’s performance on these datasets, the project aims to offer important insights into the technology’s ...