Equivariant Neural Rendering

TL;DR

This paper introduces a neural rendering framework that learns 3D scene representations from images by enforcing equivariance to 3D transformations, enabling real-time rendering without 3D supervision.

Contribution

It presents a novel equivariance-based loss for neural scene representations and introduces new datasets with complex lighting and backgrounds.

Findings

Achieves real-time scene inference and rendering.

Performs comparably to models with 3D supervision.

Successfully handles scenes with complex lighting and backgrounds.

Abstract

We propose a framework for learning neural scene representations directly from images, without 3D supervision. Our key insight is that 3D structure can be imposed by ensuring that the learned representation transforms like a real 3D scene. Specifically, we introduce a loss which enforces equivariance of the scene representation with respect to 3D transformations. Our formulation allows us to infer and render scenes in real time while achieving comparable results to models requiring minutes for inference. In addition, we introduce two challenging new datasets for scene representation and neural rendering, including scenes with complex lighting and backgrounds. Through experiments, we show that our model achieves compelling results on these datasets as well as on standard ShapeNet benchmarks.