Machine Learning in Architectural Surveying: Possibility or Next Step of Development? From Photogrammetry to Augmented Reality of a Sculptural Group

This paper investigates the future role of machine learning in architectural surveying workflows, focusing on the digitization, optimization, and dissemination of sculptural heritage through photogrammetry and augmented reality. The study is based on the Sacred Mount of San Vivaldo in Tuscany, a sixteenth-century devotional complex composed of chapels containing terracotta statuary groups. Extensive image-based surveys were carried out using high-resolution DSLR photography to generate dense photogrammetric models of the sculptures. The authors discuss how machine learning could support multiple stages of the process, including automated image acquisition, quality control, mesh simplification, texture baking, and adaptive optimization for AR platforms and 3D printing. Particular attention is given to the challenge of converting highly detailed survey models into lightweight yet accurate assets suitable for mobile visualization and public interaction. The research also proposes AR applications capable of enriching the visitor experience with historical information and interactive content. The study concludes that AI-assisted surveying can become a major next step in heritage documentation, reducing manual effort while improving accessibility, reuse, and communication of complex cultural assets.