Structure and Methodology for Building a Dataset of Images of Roman Brick Masonry

This contribution presents the construction of a dataset of images of Opus Testaceum masonry walls, structured to train supervised classification models. A total of 24 representative buildings were selected, distributed across the 14 Regions of Augustan Rome and referring to six main construction phases between the 1st and 4th centuries AD. The photographic survey produced approximately 2,400 images, varied in lighting and angle, to capture the morphological and material diversity of the masonry. The dataset is designed to test the model’s ability to generalize classification to previously unseen buildings and will be made available in open access for the benefit of the scientific community. The project also aims to foster a dialogue between the humanities and computer science, emphasizing the active role of Archaeologists and Architects in shaping innovative digital tools for the analysis, understanding, and interpretation of Cultural Heritage.