Outdoor Scene Extrapolation with Hierarchical Generative Cellular Automata

TL;DR

This paper introduces hierarchical Generative Cellular Automata (hGCA), a novel 3D generative model that extrapolates detailed street scene geometry from sparse LiDAR data, improving realism and completeness for autonomous vehicle simulations.

Contribution

The paper presents a scalable, recursive 3D generative model with a local-to-global coherence mechanism, advancing scene extrapolation beyond current methods.

Findings

hGCA outperforms state-of-the-art baselines in scene plausibility and fidelity.

The model generalizes well from synthetic to real-world data.

It can generate novel objects from limited geometric cues.

Abstract

We aim to generate fine-grained 3D geometry from large-scale sparse LiDAR scans, abundantly captured by autonomous vehicles (AV). Contrary to prior work on AV scene completion, we aim to extrapolate fine geometry from unlabeled and beyond spatial limits of LiDAR scans, taking a step towards generating realistic, high-resolution simulation-ready 3D street environments. We propose hierarchical Generative Cellular Automata (hGCA), a spatially scalable conditional 3D generative model, which grows geometry recursively with local kernels following, in a coarse-to-fine manner, equipped with a light-weight planner to induce global consistency. Experiments on synthetic scenes show that hGCA generates plausible scene geometry with higher fidelity and completeness compared to state-of-the-art baselines. Our model generalizes strongly from sim-to-real, qualitatively outperforming baselines on the Waymo-open dataset. We also show anecdotal evidence of the ability to create novel objects from real-world geometric cues even when trained on limited synthetic content. More results and details can be found on https://research.nvidia.com/labs/toronto-ai/hGCA/.