NeRF-VAE: A Geometry Aware 3D Scene Generative Model

TL;DR

NeRF-VAE is a novel 3D scene generative model that combines NeRF with VAE, enabling efficient inference and rendering of geometrically consistent scenes from limited inputs, with improved generalization and an attention mechanism.

Contribution

It introduces NeRF-VAE, integrating geometric structure into a VAE framework for 3D scene generation and inference without retraining, using differentiable volume rendering.

Findings

NeRF-VAE can infer and render new scenes from few images.

The model generalizes well to out-of-distribution camera views.

Attention-based conditioning improves model performance.

Abstract

We propose NeRF-VAE, a 3D scene generative model that incorporates geometric structure via NeRF and differentiable volume rendering. In contrast to NeRF, our model takes into account shared structure across scenes, and is able to infer the structure of a novel scene -- without the need to re-train -- using amortized inference. NeRF-VAE's explicit 3D rendering process further contrasts previous generative models with convolution-based rendering which lacks geometric structure. Our model is a VAE that learns a distribution over radiance fields by conditioning them on a latent scene representation. We show that, once trained, NeRF-VAE is able to infer and render geometrically-consistent scenes from previously unseen 3D environments using very few input images. We further demonstrate that NeRF-VAE generalizes well to out-of-distribution cameras, while convolutional models do not. Finally, we introduce and study an attention-based conditioning mechanism of NeRF-VAE's decoder, which improves model performance.