ReLaGS: Relational Language Gaussian Splatting
This addresses the problem of costly and fragmented 3D reasoning methods for researchers and practitioners in computer vision and robotics, offering a scalable solution without scene-specific training.
The paper tackles the challenge of unified 3D perception and reasoning across tasks like segmentation and relation understanding by introducing a framework that constructs a hierarchical language-distilled Gaussian scene and 3D semantic scene graph without scene-specific training, achieving efficient open-vocabulary 3D reasoning validated on tasks such as open-vocabulary segmentation and scene graph generation.
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/