Language-Embedded Gaussian Splats (LEGS): Incrementally Building Room-Scale Representations with a Mobile Robot

TL;DR

LEGS is a novel 3D scene mapping system that incrementally builds detailed, semantic-rich representations using a mobile robot, enabling efficient open-vocabulary object localization in indoor environments.

Contribution

The paper introduces LEGS, a new online training method for semantic 3D mapping that is faster and effective for localizing diverse objects in indoor scenes.

Findings

LEGS trains over 3.5x faster than LERF.

LEGS achieves up to 66% accuracy in object localization.

Multi-camera setup improves visual reconstruction quality.

Abstract

Building semantic 3D maps is valuable for searching for objects of interest in offices, warehouses, stores, and homes. We present a mapping system that incrementally builds a Language-Embedded Gaussian Splat (LEGS): a detailed 3D scene representation that encodes both appearance and semantics in a unified representation. LEGS is trained online as a robot traverses its environment to enable localization of open-vocabulary object queries. We evaluate LEGS on 4 room-scale scenes where we query for objects in the scene to assess how LEGS can capture semantic meaning. We compare LEGS to LERF and find that while both systems have comparable object query success rates, LEGS trains over 3.5x faster than LERF. Results suggest that a multi-camera setup and incremental bundle adjustment can boost visual reconstruction quality in constrained robot trajectories, and suggest LEGS can localize open-vocabulary and long-tail object queries with up to 66% accuracy.