How to compute a shape : optical character recognition for hieratic

Abstract

This paper is written in the framework of the project Altägyptische Kursivschriften and presents an experiment for applying OCR to Hieratic, examining the final step of the OCR pipeline. A convolutional neural network is used to classify individual hieratic characters. The results prove that the classification of hieratic characters is in principle possible, but they also show, where improvements and combinations with other methods promise a better performance of recognition models.