Editorial Vol. 16

Artificial intelligence and digitalization of cultural heritage: new frontiers for architectural and museum research

In recent years, the field of research on Cultural Heritage and architecture has undergone a radical transformation, driven by the progressive integration of advanced digital modeling tools and artificial intelligence. The contributions gathered in this issue of the journal aim to explore some of the trajectories outlined by this transformation, showing how the adoption of advanced and integrated approaches is redefining methods of representation, analysis, and management of cultural heritage. 

A first line of inquiry concerns applications related to architectural representation. The use of generative models based on diffusion-based techniques, such as Stable Diffusion, now makes it possible to experiment with stylistic transposition as a process of critical learning. Through fine-tuning procedures carried out with DreamBooth and orchestrated via nodal pipelines in ComfyUI, controlled by tools such as ControlNet, it is possible to transmit to AI models the graphic codes of figures such as Paolo Portoghesi, Ludwig Mies van der Rohe, and Le Corbusier. The generated results—coherent yet not imitative—demonstrate how AI, when properly guided, can act not only as an executive tool but as a critical agent capable of revealing latent structures of architectural style, opening the way to new forms of representation. 

A second research trajectory emerging from the articles presented here concerns the significant evolution of immersive visualization and spatial documentation techniques. One case involves the integration of spherical panoramas within Building Information Modeling (BIM) models, enabling more effective engagement with museum spaces. Alongside this, the use of Structure from Motion (SfM) pipelines such as GLOMAP, enhanced by AI-based approaches like Neural Radiance Fields (NeRF) and 3D Gaussian Splatting, allows the reconstruction of three-dimensional scenes even from low-resolution monocular or 360° video, improving both efficiency and quality of models for the purposes of heritage conservation and enhancement. 

Another field of experimentation involves the creation of datasets specifically designed for the classification of historical construction techniques, with the goal of training dedicated AI mod

els. In this context, the systematic collection of images of Opus Testaceum masonry is aimed not only at training supervised classification models but also at fostering a structured and interdisciplinary dialogue between archaeologists and computer scientists. This approach seeks to develop models capable of recognizing construction techniques even in buildings not included in the training dataset, thereby strengthening their generalization capacity. At the same time, the creation of open-access datasets represents a concrete contribution to the scientific community, encouraging further research and applications. 

In the field of conservation, experiments are emerging that combine Heritage Building Information Modeling (H-BIM) with convolutional neural networks for structural damage analysis. Case studies include historic churches such as the Church of the Compagnia della Disciplina della Santa Croce in Naples. The integration of crack-related information into semantic models makes it possible to move beyond mere geometric description, paving the way for semi-automatic risk assessments and interoperable information systems for the preventive management of cultural heritage. 

Finally, the theme of digitizing museum collections is addressed within the framework of the project MUSE – MUseum management, enhancement and accessibility: a Sustainable digital Ecosystem of interactive digital collections, with an experiment conducted at the Galleria Nazionale d’Arte Antica in Palazzo Barberini, Rome. Here, the construction of digital twins is understood as a critical process of copying, based on principles of data neutrality and process transparency, in accordance with the National Plan for the Digitalization of Cultural Heritage (PND). The objective is to define reliability parameters for digital replicas and promote a sustainable, interactive ecosystem of digital collections. 

Taken together, these contributions testify to a paradigm shift: artificial intelligence and digital methods are no longer mere auxiliary tools but active and generative components of the processes of knowledge, representation, and heritage management. They presuppose a critical and conscious preparation on the part of users. The hybridization of humanistic knowledge and computational technologies thus opens unprecedented scenarios, inviting us to critically rethink and relaunch the roles, languages, and responsibilities of contemporary research.