EGYPTIAN ARTIFACTS’ MATERIALS CLASSIFIER BASED ON LIGHTWEIGHT DEEP LEARNING ARCHITECTURES

M., Esmat,; K., Moussa,; R., Yousri,; S., Alwardany,; M., Wessam,; H., Mostafa,; M., Darweesh,

doi:10.21608/ejars.2025.434903

EGYPTIAN ARTIFACTS’ MATERIALS CLASSIFIER BASED ON LIGHTWEIGHT DEEP LEARNING ARCHITECTURES

Document Type : Original Article

Authors

¹ School of Information Technology & Computer Science, Nile Univ., Giza, Egypt.

² Wireless Intelligent Networks Center (WINC), Nile Univ., Giza, Egypt. School of Engineering & Applied Sciences, Nile Univ., Giza, Egypt.

³ Wireless Intelligent Networks Center (WINC), Nile Univ., Giza, Egypt.

⁴ School of Engineering & Applied Sciences, Nile Univ., Giza, Egypt.

⁵ Electronics & Communications Engineering dept., Cairo Univ, Giza, Egypt. Nanotechnology & Nanoelectronics Program, University of Science and Technology, Zewail City, Giza, Egypt.

10.21608/ejars.2025.434903

Abstract

Artificial Intelligence (AI) plays a crucial role in cultural heritage by enabling the analysis, preservation, and restoration of artifacts and historical documents. Most of these applications may require to be used on devices with limited resources which leads to the need to use lightweight models. This study employs lightweight deep learning models, MobileNet V3 and ResNet-50, to classify Egyptian artifacts based on seven different materials. The models are trained on a dataset of 10.274 images. MobileNet achieves a training accuracy of 99.6% and a validation accuracy of 78.75%, while ResNet-50 achieves 96.62% and 83.23%, respectively. This research represents a novel contribution as previous studies have not specifically add-ressed the classification of materials in Egyptian artifacts. Such advancements highlight AI's potential in making cultural heritage more accessible and enhancing historical under-sta-nding.

Keywords