Implicit neural representation of textures

TL;DR

This paper introduces a novel implicit neural representation for textures that operates continuously over UV space, demonstrating high image quality and exploring applications in real-time rendering and downstream tasks.

Contribution

It proposes a new texture INR design that functions continuously over UV coordinates, with comprehensive analysis and application exploration.

Findings

High-quality texture rendering achieved

Memory and inference time analyzed and balanced

Effective in real-time rendering and downstream tasks

Abstract

Implicit neural representation (INR) has proven to be accurate and efficient in various domains. In this work, we explore how different neural networks can be designed as a new texture INR, which operates in a continuous manner rather than a discrete one over the input UV coordinate space. Through thorough experiments, we demonstrate that these INRs perform well in terms of image quality, with considerable memory usage and rendering inference time. We analyze the balance between these objectives. In addition, we investigate various related applications in real-time rendering and down-stream tasks, e.g. mipmap fitting and INR-space generation.