Exploring the Application of Interpretable Neural Networks for the Petrographic Classification of Ceramic Samples from the Levant |
|---|
| Sara Capriotti, Alessio Devoto, Donatella Genovese, Silvano Mignardi, Simone Scardapane, Laura Medeghini |
- Abstract:
- The archaeological context of the Levantine region is both rich and complex, particularly during the transition from the Late Chalcolithic to the Early Bronze Age, a period marked by urban development, craft specialization, and interregional trade. This study explores the use of Artificial Intelligence techniques to classify Levantine ceramic thin sections based on their petrographic fabrics. Deep learning models, including Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs), were applied to a large dataset of thin section images from ceramics dating to the Uruk period, Bronze Age, and Iron Age, collected from various archaeological sites across the Levant. To improve model transparency, explainable AI methods such as Guided Grad-CAM and attention maps were applied to identify key features and interpret latent representations. The results show that deep learning can achieve high accuracy in automated ceramic classification and provide important insights into ancient ceramic technologies and cultural interactions.
- Download:
- IMEKO-Metroarchaeo-2025-045.pdf
- DOI:
- 10.21014/tc26-2025.045
- Event details
- IMEKO TC:
- TC26
- Event name:
- TC26 MetroArcheo Conference 2025
- Title:
Metrology for Archaeology and Cultural Heritage
- Place:
- Bergamo, ITALY
- Time:
- 15 October 2025 - 17 October 2025