Mo.Se.: Mosaic image segmentation based on deep cascading learning

A novel algorithm, Mo.Se. (Mosaic Segmentation), significantly enhances the accuracy of automatically detecting tesserae in ancient mosaics by combining deep learning with image segmentation techniques. This approach, which integrates U-Net 3 Network and the Watershed algorithm, outperforms existing methods by nearly 10% in overall accuracy, as validated using publicly available datasets. Tested on a high-resolution orthoimage of a mosaic from St. Stephen's Church at Um er-Rasas (Kastrom Mefa'a), a UNESCO World Heritage Site in Jordan, the method provides a scalable and efficient solution for digitizing mosaics, reducing the reliance on manual and time-consuming processes. This advancement offers archaeologists a robust tool for studying, comparing, and preserving ancient geometric mosaics.

Felicetti, A., Paolanti, M., Zingaretti, P., Pierdicca, R., & Malinverni, E. S. (2021). Mo.Se.: Mosaic image segmentation based on deep cascading learning. Virtual Archaeology Review, 12(24), 25. https://doi.org/10.4995/var.2021.14179