Plenoxels: Radiance Fields without Neural Networks

TL;DR

Plenoxels is a neural network-free, sparse voxel-based system for photorealistic view synthesis that is significantly faster to optimize than Neural Radiance Fields without sacrificing visual quality.

Contribution

It introduces a neural network-free, sparse voxel representation for view synthesis that is optimized efficiently using gradient methods and regularization.

Findings

Optimized two orders of magnitude faster than Neural Radiance Fields.

Achieves comparable visual quality to neural network-based methods.

Effective on standard benchmark tasks.

Abstract

We introduce Plenoxels (plenoptic voxels), a system for photorealistic view synthesis. Plenoxels represent a scene as a sparse 3D grid with spherical harmonics. This representation can be optimized from calibrated images via gradient methods and regularization without any neural components. On standard, benchmark tasks, Plenoxels are optimized two orders of magnitude faster than Neural Radiance Fields with no loss in visual quality.