Outside the wall: some applications of TagLab for semantic segmentation in archaeological practice

Imaging techniques have a long history in archeology, from the ruins depicted during the Grand Tour to today's large-scale photogrammetric surveys. Orthophotos generated with photogrammetry are currently extensively used in archeology: a key challenge in this new documentation paradigm lies in the semantic segmentation of such pixel matrices, a process increasingly aided by Machine Learning. This paper focuses on TagLab, an open-source tool originally developed for marine biology that has proven to be extremely versatile in archeological practice. More specifically, this paper aims to expand the use of the software in the Cultural Heritage field and explore its applicability for cost estimation on ancient construction sites, archeological excavations, ceramic thin sections and fragmentary painting documentation.