L3DG: Latent 3D Gaussian Diffusion

TL;DR

L3DG introduces a novel latent diffusion method for 3D Gaussian modeling, enabling efficient, high-detail, room-scale scene generation with real-time rendering capabilities.

Contribution

It is the first to apply latent 3D Gaussian diffusion for scalable, detailed scene synthesis using a VQ-VAE and sparse convolutions.

Findings

Improves visual quality over prior object-level synthesis methods.

Enables real-time rendering of generated scenes from arbitrary viewpoints.

Scales effectively to room-scale scenes with high detail.

Abstract

We propose L3DG, the first approach for generative 3D modeling of 3D Gaussians through a latent 3D Gaussian diffusion formulation. This enables effective generative 3D modeling, scaling to generation of entire room-scale scenes which can be very efficiently rendered. To enable effective synthesis of 3D Gaussians, we propose a latent diffusion formulation, operating in a compressed latent space of 3D Gaussians. This compressed latent space is learned by a vector-quantized variational autoencoder (VQ-VAE), for which we employ a sparse convolutional architecture to efficiently operate on room-scale scenes. This way, the complexity of the costly generation process via diffusion is substantially reduced, allowing higher detail on object-level generation, as well as scalability to large scenes. By leveraging the 3D Gaussian representation, the generated scenes can be rendered from arbitrary viewpoints in real-time. We demonstrate that our approach significantly improves visual quality over prior work on unconditional object-level radiance field synthesis and showcase its applicability to room-scale scene generation.