Abstract: In the past, archaeologists have attempted to use various software tools to facilitate the study and digitization of ancient languages, with varying degrees of success. The aim of this study is to develop an object detection program that is trained to recognize the syllabary of Linear-A, a Minoan (Cretan) script of the Bronze Age. Specifically, this study will focus on the ‘T’ series, which consists of 6 syllabograms, to determine if the use of this particular program can feasibly produce acceptable results for our case’s requirements. While an optical character recognition (OCR) program was initially suggested, its performance would be sub-optimal in our case (which consists mostly of clay tablet images), so it was decided to use an object detection program instead. YOLOv5 was chosen for its low hardware requirements, quick speed and high accuracy, while also being easy to use. The study concluded that the YOLOv5 program successfully recognized the syllabograms, with a satisfactory success rate. Nevertheless, there is always room for improvement in the results or the use of a newer version of the program.

Keywords: Linear-A Script, Bronze Age Crete, Clay Tablets, Machine Vision, Object Detection, YOLOv5

Pages: 16-21