Tarsier: Evolving Noise Injection in Super-Resolution GANs

TL;DR

This paper introduces Tarsier, an evolutionary approach to optimize noise injection in super-resolution GANs at inference time, improving image quality and realism over existing methods like NESRGAN+.

Contribution

It presents a novel evolutionary optimization framework for noise injection in super-resolution GANs, enhancing image quality without retraining the model.

Findings

Outperforms NESRGAN+ on standard datasets

Improves perceptual quality according to PIRM score

Validated by human study

Abstract

Super-resolution aims at increasing the resolution and level of detail within an image. The current state of the art in general single-image super-resolution is held by NESRGAN+, which injects a Gaussian noise after each residual layer at training time. In this paper, we harness evolutionary methods to improve NESRGAN+ by optimizing the noise injection at inference time. More precisely, we use Diagonal CMA to optimize the injected noise according to a novel criterion combining quality assessment and realism. Our results are validated by the PIRM perceptual score and a human study. Our method outperforms NESRGAN+ on several standard super-resolution datasets. More generally, our approach can be used to optimize any method based on noise injection.