Internship Project Cuneiform Detector

   

Development of the Cuneiform Detector, a web application for the semantic segmentation of cuneiform tablets using deep learning.

  

This project developed a semantic segmentation model for photographs of cuneiform tablets. The model uses a convolutional neural network to separate the text from any sort of background, including broken surfaces. Interestingly enough, the model works to some extend on older stages of Akkadian and even Sumerian tablets, despite being mainly trained on Neo-Babylonian data.