Furthering 3D Digital Representation Methods: An Introduction to the Application of Neural Radiance Fields as an Alternative to Photogrammetric Modeling

Summary

This is an abstract from the "SAA 2024: Individual Abstracts" session, at the 89th annual meeting of the Society for American Archaeology.

Photogrammetry has seen increasing utilization within archaeology in recent years but with the rise of this representational methodology has come several challenges including the loss of context, inaccurate reproduction of surfaces, and difficulties processing thin objects. Emerging free open-source machine learning technology can produce novel scenes known as Neural Radiance Fields (NeRFs), which present a valuable alternative when creating 3D representations of objects, sites, and survey areas. While photogrammetry reproduces 3D geometry by matching features to create a polygonal mesh, the artificial intelligence powered NeRF software creates a continuous scene made of view-dependent radiating sources of light. This allows NeRFs to depict elements with unclear geometry, such as the sky and a more accurate representation of how light reflects off the subject from different angles; therefore, producing more lifelike facsimiles of the subject matter. NeRF utilizes the same data collection methods as photogrammetry, making it an accessible addition to the archaeologist’s digital repertoire that can be employed not only on future endeavors but past ones as well. The research presented demonstrates the various facets of NeRF technology and how it can help us preserve, record, analyze, and educate within an archaeological context.

This Resource is Part of the Following Collections

• SAA 2024: Individual Abstracts   •
• Society for American Archaeology 89th Annual Meeting, New Orleans, LA (2024)

Keywords

Record Identifiers

Abstract Id(s): 41503.0