QFF: Quantized Fourier Features for Neural Field Representations

TL;DR

This paper introduces Quantized Fourier Features (QFF), a novel encoding method that improves neural field representations by enabling faster training, smaller models, and higher quality outputs across various applications.

Contribution

QFF provides a multiresolution, periodic encoding that enhances neural field models without increasing complexity, serving as a simple and effective drop-in replacement.

Findings

QFF reduces model size and training time.

QFF improves output quality in neural image and 3D representations.

QFF is easy to implement and computationally efficient.

Abstract

Multilayer perceptrons (MLPs) learn high frequencies slowly. Recent approaches encode features in spatial bins to improve speed of learning details, but at the cost of larger model size and loss of continuity. Instead, we propose to encode features in bins of Fourier features that are commonly used for positional encoding. We call these Quantized Fourier Features (QFF). As a naturally multiresolution and periodic representation, our experiments show that using QFF can result in smaller model size, faster training, and better quality outputs for several applications, including Neural Image Representations (NIR), Neural Radiance Field (NeRF) and Signed Distance Function (SDF) modeling. QFF are easy to code, fast to compute, and serve as a simple drop-in addition to many neural field representations.