arXiv · 2611.10644
Agent2BIM: A Systematic Evaluation of Reality-Capture Configurations and Prior Models for As-Built Modeling
Abstract
AI agents can generate as-built Building Information Models (BIMs) from reality-capture data, but the evidence required for accurate reconstruction remains unclear. This paper presents Agent2BIM, an agent-based Scan-to-BIM framework, and evaluates seven capture configurations in a research laboratory and a mechanical, electrical, and plumbing (MEP) scene. Inputs include 360° panoramas, smartphone video, smartphone LiDAR, terrestrial laser scanning (TLS), TLS with posed panoramas, and imagery supplemented by prior IFC models. Outputs are compared against TLS-verified reference models for component detection, geometric accuracy, and MEP connectivity. In the laboratory, component recall ranged from 84% to 100%, but only 9% of components reconstructed from panoramas were within 20 cm of their reference positions, compared with 86% using smartphone LiDAR. In the MEP scene, scale-free inputs recovered 11 - 43% of components, versus 75 - 93% for metric inputs. Prior IFC models improved MEP reconstruction but could propagate inaccuracies. The results highlight the complementary roles of visual and metric evidence in agent-based Scan-to-BIM.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Boyu Wang, Borja García de Soto. 2026-11-07. Agent2BIM: A Systematic Evaluation of Reality-Capture Configurations and Prior Models for As-Built Modeling. https://arxiv.science/abs/2611.10644
Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite arXiv Science by Indiciae: BibTeX · RIS.