We are pleased to announce the publication of the new paper titled “Automatic Removal of Non-Architectural Elements in 3D Models of Historic Buildings with Language Embedded Radiance Fields” in the journal Heritage. This research is the outcome of a collaborative effort between EPFL’s Earthquake Engineering and Structural Dynamics Laboratory (EESD) and Digital Humanities Laboratory (DHLab), with support from the EPFL Center for Imaging.
Authors: Alexander Rusnak (Digital Humanities Laboratory, EPFL), Bryan G. Pantoja-Rosero (Earthquake Engineering and Structural Dynamics Laboratory, EPFL), Frédéric Kaplan (Digital Humanities Laboratory, EPFL), Katrin Beyer ((Earthquake Engineering and Structural Dynamics Laboratory, EPFL)
This paper is integral to the “3D Imaging of Historical Buildings” project, which aims to advance the preservation of cultural heritage through cutting-edge 3D modeling methodologies. Our research introduces a novel pipeline leveraging neural radiance fields (NeRF) to automatically generate purified architectural structure-only point clouds. This technique markedly improves the assessment of structural integrity in historic buildings by excising non-architectural elements, thereby ensuring privacy and enhancing the precision and accuracy of the models.