Machine Learning in Archaeological Shape Analysis

2025 | Jerusalem | Ongoing research

Building on The Riddle of the Cross, this study applies machine learning to the analysis of engraved crosses at archaeological sites. Classifying such marks has traditionally relied on experts sorting them by hand into typologies, a slow and partly subjective process. The project develops a semi-automated pipeline instead: preprocessing of 3D data, detection of crosses in 2D and 3D, segmentation from the stone surface, feature extraction, and clustering to identify shape prototypes, supported by two dedicated interfaces for analysis and presentation.

The preliminary results point toward scalable, objective and reproducible methods for shape classification and typological study in archaeology.

Links

• The Riddle of the Cross