BIM for built cultural heritage. Semantic segmentation, architectural stratification and LOD of the Baptistery of San Giovanni in Florence.

The 3D virtual modeling of any artifact involves the organized composition of digital elements. This process is common to any 3D modeling workflow but, when applied to BIM systems, shows a good correspondence in the construction processes between the virtual and the real world. In fact, any building can be considered as a coordinated set of simple technological elements, linked together by design logics and construction techniques that are quite predictable and controllable. Even existing buildings can be "deconstructed" into recurring elements, especially if the architectural survey operations reveal the regularity of certain geometric patterns; however, parametric and informative modeling of built artifacts is much more difficult, both in terms of geometric transposition of the continuity of the real world and of its qualitative and semantic description. These difficulties are also associated with the intrinsic rigidity of the parametric modeling workflow, subjected to "libraries" of digital objects that clash against the variability and uniqueness of the built environment, especially when "historic" or in a poor state of conservation. While the transition from numeric to geometric models involves a crucial critical "discretization" operation, in the construction of the BIM, the model involves a further interpretative step: the semantic structuring of its compositional elements. In this framework, the presented study becomes an application opportunity for the theoretical issues highlighted above, using the Baptistery of San Giovanni in Florence as a case study. The modeling of this monument, relevant from a historical point of view and for its particular stylistic and geometric features, is based on a massive, integrated survey conducted by the University of San Diego. Specifically, we intend to analyze not only how to discern the constructive elements through their semantic segmentation, but above all how to connect them and make them interact with each other. As a result, we aim at providing a representation of the different evolutionary phases and above all propose a more effective meaning for the Levels of Development parameter when applied to built heritage.