Paper ID: 2411.03540

VLA-3D: A Dataset for 3D Semantic Scene Understanding and Navigation

Haochen Zhang, Nader Zantout, Pujith Kachana, Zongyuan Wu, Ji Zhang, Wenshan Wang

With the recent rise of Large Language Models (LLMs), Vision-Language Models (VLMs), and other general foundation models, there is growing potential for multimodal, multi-task embodied agents that can operate in diverse environments given only natural language as input. One such application area is indoor navigation using natural language instructions. However, despite recent progress, this problem remains challenging due to the spatial reasoning and semantic understanding required, particularly in arbitrary scenes that may contain many objects belonging to fine-grained classes. To address this challenge, we curate the largest real-world dataset for Vision and Language-guided Action in 3D Scenes (VLA-3D), consisting of over 11.5K scanned 3D indoor rooms from existing datasets, 23.5M heuristically generated semantic relations between objects, and 9.7M synthetically generated referential statements. Our dataset consists of processed 3D point clouds, semantic object and room annotations, scene graphs, navigable free space annotations, and referential language statements that specifically focus on view-independent spatial relations for disambiguating objects. The goal of these features is to aid the downstream task of navigation, especially on real-world systems where some level of robustness must be guaranteed in an open world of changing scenes and imperfect language. We benchmark our dataset with current state-of-the-art models to obtain a performance baseline. All code to generate and visualize the dataset is publicly released, see this https URL With the release of this dataset, we hope to provide a resource for progress in semantic 3D scene understanding that is robust to changes and one which will aid the development of interactive indoor navigation systems.

Submitted: Nov 5, 2024