ShelfGaussian: Shelf-Supervised Open-Vocabulary Gaussian-based 3D Scene Understanding

TL;DR

ShelfGaussian is a novel 3D scene understanding framework that leverages off-the-shelf vision foundation models to enable open-vocabulary, multi-modal Gaussian representations for improved perception and planning in diverse environments.

Contribution

It introduces a Multi-Modal Gaussian Transformer and a Shelf-Supervised Learning Paradigm to optimize Gaussian representations across multiple sensor modalities and scene levels.

Findings

Achieves state-of-the-art zero-shot semantic occupancy prediction on Occ3D-nuScenes.

Demonstrates effective in-the-wild performance on urban scenarios with UGVs.

Abstract

We introduce ShelfGaussian, an open-vocabulary multi-modal Gaussian-based 3D scene understanding framework supervised by off-the-shelf vision foundation models (VFMs). Gaussian-based methods have demonstrated superior performance and computational efficiency across a wide range of scene understanding tasks. However, existing methods either model objects as closed-set semantic Gaussians supervised by annotated 3D labels, neglecting their rendering ability, or learn open-set Gaussian representations via purely 2D self-supervision, leading to degraded geometry and limited to camera-only settings. To fully exploit the potential of Gaussians, we propose a Multi-Modal Gaussian Transformer that enables Gaussians to query features from diverse sensor modalities, and a Shelf-Supervised Learning Paradigm that efficiently optimizes Gaussians with VFM features jointly at 2D image and 3D scene levels. We evaluate ShelfGaussian on various perception and planning tasks. Experiments on Occ3D-nuScenes demonstrate its state-of-the-art zero-shot semantic occupancy prediction performance. ShelfGaussian is further evaluated on an unmanned ground vehicle (UGV) to assess its in the-wild performance across diverse urban scenarios. Project website: https://lunarlab-gatech.github.io/ShelfGaussian/.