Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

This paper introduces a new approach to super-resolution (SR), the process of enhancing image resolution, using Generative Adversarial Networks (GAN). The SRGAN model is trained to generate high-resolution images that are not only visually pleasing but also more realistic by using a perceptual loss function and an adversarial network. This results in images with finer details compared to traditional SR methods, making the upscaled images more convincing and closer to real-world scenes.