SceneDreamer: Unbounded 3D Scene Generation from 2D Image Collections

TL;DR

SceneDreamer introduces a novel generative model capable of creating large-scale, unbounded 3D landscapes from 2D image collections without requiring 3D annotations, using an efficient scene representation and neural rendering.

Contribution

It proposes a new framework combining a BEV scene representation, neural hash grid, and neural volumetric renderer for unbounded 3D scene generation from 2D images.

Findings

Outperforms state-of-the-art in 3D scene generation quality

Efficient training with quadratic complexity

Generates diverse and photorealistic 3D worlds

Abstract

In this work, we present SceneDreamer, an unconditional generative model for unbounded 3D scenes, which synthesizes large-scale 3D landscapes from random noise. Our framework is learned from in-the-wild 2D image collections only, without any 3D annotations. At the core of SceneDreamer is a principled learning paradigm comprising 1) an efficient yet expressive 3D scene representation, 2) a generative scene parameterization, and 3) an effective renderer that can leverage the knowledge from 2D images. Our approach begins with an efficient bird's-eye-view (BEV) representation generated from simplex noise, which includes a height field for surface elevation and a semantic field for detailed scene semantics. This BEV scene representation enables 1) representing a 3D scene with quadratic complexity, 2) disentangled geometry and semantics, and 3) efficient training. Moreover, we propose a novel generative neural hash grid to parameterize the latent space based on 3D positions and scene semantics, aiming to encode generalizable features across various scenes. Lastly, a neural volumetric renderer, learned from 2D image collections through adversarial training, is employed to produce photorealistic images. Extensive experiments demonstrate the effectiveness of SceneDreamer and superiority over state-of-the-art methods in generating vivid yet diverse unbounded 3D worlds.