ReLaGS: Relational Language Gaussian Splatting

TL;DR

ReLaGS introduces a novel framework for 3D scene understanding that constructs hierarchical language-distilled Gaussian scenes and 3D semantic scene graphs without scene-specific training, enabling efficient open-vocabulary reasoning across multiple tasks.

Contribution

It presents a new method combining Gaussian scene modeling, language alignment, and graph reasoning for scalable 3D perception without scene-specific training.

Findings

Effective open-vocabulary 3D segmentation

Accurate 3D scene graph generation

Relational reasoning across tasks

Abstract

Achieving unified 3D perception and reasoning across tasks such as segmentation, retrieval, and relation understanding remains challenging, as existing methods are either object-centric or rely on costly training for inter-object reasoning. We present a novel framework that constructs a hierarchical language-distilled Gaussian scene and its 3D semantic scene graph without scene-specific training. A Gaussian pruning mechanism refines scene geometry, while a robust multi-view language alignment strategy aggregates noisy 2D features into accurate 3D object embeddings. On top of this hierarchy, we build an open-vocabulary 3D scene graph with Vision Language derived annotations and Graph Neural Network-based relational reasoning. Our approach enables efficient and scalable open-vocabulary 3D reasoning by jointly modeling hierarchical semantics and inter/intra-object relationships, validated across tasks including open-vocabulary segmentation, scene graph generation, and relation-guided retrieval. Project page: https://dfki-av.github.io/ReLaGS/