Variable Bitrate Neural Fields

TL;DR

This paper introduces a dictionary-based compression method for feature grids in neural fields, significantly reducing memory usage and enabling multiresolution representations for efficient streaming.

Contribution

It proposes a vector-quantized auto-decoder approach to learn discrete neural representations, improving memory efficiency of neural field models.

Findings

Memory consumption reduced by up to 100x

Supports multiresolution and out-of-core streaming

Enables end-to-end learning of discrete neural representations

Abstract

Neural approximations of scalar and vector fields, such as signed distance functions and radiance fields, have emerged as accurate, high-quality representations. State-of-the-art results are obtained by conditioning a neural approximation with a lookup from trainable feature grids that take on part of the learning task and allow for smaller, more efficient neural networks. Unfortunately, these feature grids usually come at the cost of significantly increased memory consumption compared to stand-alone neural network models. We present a dictionary method for compressing such feature grids, reducing their memory consumption by up to 100x and permitting a multiresolution representation which can be useful for out-of-core streaming. We formulate the dictionary optimization as a vector-quantized auto-decoder problem which lets us learn end-to-end discrete neural representations in a space where no direct supervision is available and with dynamic topology and structure. Our source code will be available at https://github.com/nv-tlabs/vqad.