Paired and Unpaired Image to Image Translation using Generative Adversarial Networks

TL;DR

This paper explores paired and unpaired image-to-image translation using GANs, analyzing various architectures, loss functions, and metrics to improve and evaluate translation quality across multiple domains.

Contribution

It introduces a comprehensive study of both paired and unpaired image translation with new quantitative metrics and detailed experimental analysis.

Findings

Cycle consistency loss improves unpaired translation quality.

Different Patch-GAN sizes affect the realism of generated images.

New metrics like precision, recall, and FID provide better evaluation.

Abstract

Image to image translation is an active area of research in the field of computer vision, enabling the generation of new images with different styles, textures, or resolutions while preserving their characteristic properties. Recent architectures leverage Generative Adversarial Networks (GANs) to transform input images from one domain to another. In this work, we focus on the study of both paired and unpaired image translation across multiple image domains. For the paired task, we used a conditional GAN model, and for the unpaired task, we trained it using cycle consistency loss. We experimented with different types of loss functions, multiple Patch-GAN sizes, and model architectures. New quantitative metrics - precision, recall, and FID score - were used for analysis. In addition, a qualitative study of the results of different experiments was conducted.