Differentiable Point-Based Radiance Fields for Efficient View Synthesis

TL;DR

This paper introduces a fast, memory-efficient differentiable point-based rendering method for novel view synthesis, significantly outperforming existing volume-based approaches in speed and resource usage with minimal quality loss.

Contribution

It presents a novel point-based representation and differentiable renderer that drastically improves speed and memory efficiency over prior volume-based methods for view synthesis.

Findings

Up to 300x faster than NeRF in training and inference.

Uses less than 10MB memory for static scenes.

Achieves near real-time rendering for dynamic scenes with high quality.

Abstract

We propose a differentiable rendering algorithm for efficient novel view synthesis. By departing from volume-based representations in favor of a learned point representation, we improve on existing methods more than an order of magnitude in memory and runtime, both in training and inference. The method begins with a uniformly-sampled random point cloud and learns per-point position and view-dependent appearance, using a differentiable splat-based renderer to evolve the model to match a set of input images. Our method is up to 300x faster than NeRF in both training and inference, with only a marginal sacrifice in quality, while using less than 10~MB of memory for a static scene. For dynamic scenes, our method trains two orders of magnitude faster than STNeRF and renders at near interactive rate, while maintaining high image quality and temporal coherence even without imposing any temporal-coherency regularizers.