Get Started
Home
Topics
Search
Library
Research questionHow reliably can open-set vision-language 3D scene graphs support outdoor navigation and object retrieval?Outdoor maps must connect open-set semantic features with geometry that supports movement and retrieval. Repeated traversals reveal whether graph semantics, routes, and region-level interpretations remain stable as an environment is revisited.
Computer Vision
Evaluation & Benchmarks
Multimodal Models
Research Paper
Robotics
Latest papersRecent research connected to this question, newest first.Open-Set 3D Scene Graphs for Field Robotics: An Outdoor Case StudyThe source analyzes Terra 3DSG across diverse outdoor robotic datasets, examining point embeddings, place-node navigation, region understanding, memory use, and consistency across repeated traversals. It is a field case study of one graph representation and tested datasets, so it does not establish general outdoor deployment performance.research paper · Sep 4, 2026
Related questions
How can fixed 3D Gaussian-splatting reconstructions support valid, traversable, closed-loop visual-navigation evaluation?How can vision-language models infer 3D geometry and temporal continuity from 2D visual observations?How can one 3D semantic occupancy model handle indoor and outdoor scenes with varying cameras, scales, and label taxonomies?How can robots maintain geometrically grounded semantic maps as objects and spatial relations change in real time?