Open-World 3D Vision-Language Scene Representation with Gaussian Splatting and Cross-Modal Alignment
Journal of the Artificial Intelligence Engineers Consortium · 2026
DOI: 10.66238/jaiec72 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of the Artificial Intelligence Engineers Consortium, 2026. Verfasst von Hikari Fujimoto. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The intersection of three-dimensional scene representation and natural language understanding represents a critical frontier in computer vision and artificial intelligence. This paper introduces a comprehensive framework for open-world three-dimensional vision-language scene representation, leveraging the computational efficiency and high-fidelity rendering capabilities of three-dimensional Gaussian Splatting combined with advanced cross-modal alignment techniques. Traditional methods relying on Neural Radiance Fields often struggle with slow rendering speeds and limited adaptability when integrating high-dimensional language features. By mapping pre-trained vision-language foundational model embeddings directly into a continuous three-dimensional Gaussian field, our approach enables real-time, open-vocabulary querying and scene interaction. We propose a novel feature distillation mechanism that aligns two-dimensional semantic maps with three-dimensional spatial structures, ensuring that nuanced textual queries accurately localize complex objects in diverse environments. Extensive theoretical analysis and simulated experimental evaluations demonstrate that the proposed architecture not only significantly reduces computational overhead but also achieves unprecedented accuracy in zero-shot segmentation and object retrieval tasks. The results validate the efficacy of utilizing explicit volumetric representations for multimodal spatial reasoning, paving the way for advanced applications in embodied artificial intelligence, robotic navigation, and interactive virtual environments
Zitieren
APA
Fujimoto, H. (2026). Open-World 3D Vision-Language Scene Representation with Gaussian Splatting and Cross-Modal Alignment. Journal of the Artificial Intelligence Engineers Consortium. https://doi.org/10.66238/jaiec72
Vancouver
Fujimoto H. Open-World 3D Vision-Language Scene Representation with Gaussian Splatting and Cross-Modal Alignment. Journal of the Artificial Intelligence Engineers Consortium. 2026. doi:10.66238/jaiec72
BibTeX
@article{Fujimoto2026,
author = {Fujimoto, Hikari},
title = {Open-World 3D Vision-Language Scene Representation with Gaussian Splatting and Cross-Modal Alignment},
journal = {Journal of the Artificial Intelligence Engineers Consortium},
year = {2026},
doi = {10.66238/jaiec72}
}
EUVIMED