Underwater Detection of Ancient Pottery Sherds Using Deep Learning


Konstantinos Paraskevas, Ioannis Mariolis, Georgios Giouvanis, GeorgiaPeleka, Georgios Zampokas and Dimitrios Tzovaras, Information Technologies Institute, Greece


This paper outlines the creation of a machine learning model designed to identify ancient pottery fragments near a submerged shipwreck of Modi Island, Greece. We trained multiple iterations of the YOLOv8 model using a custom dataset comprised of underwater videos taken during diving expeditions at the wreck site. The primary goal of this research is to integrate the resulting object detection system into a remotely operated vehicle (ROV) for automated pottery shard recognition, thereby aiding archaeological excavations. The paper elaborates on the model's development methodology and presents comprehensive experimental and evaluative results. These findings underscore the model's potential to significantly enhance the ef iciency and accuracy of underwater archaeological exploration and analysis.


Ancient pottery shreds detection, underwater archaeological excavations, machine learning, object detection, remotely operated vehicle (ROV), underwater shipwrecks, YOLOv8 model.