Project Akhet is designed to utilize the latest AI methods to recognize Ancient Egyptian texts

Contact: humonen@airi.net

Current information and updates: https://imak.ai/projects/ancient-egypt

Models

• nllb-translation/ — Middle Egyptian (hieroglyphs/transliteration) to English/German translation, weights on Hugging Face.
• pipeline/: the full reading pipeline from a page photo. Text lines, hieroglyph detection, reading direction and modern text, Gardiner sign classification, reading order as Manuel de Codage, then transliteration and translation with the model above, correctable by hand at every step. Runs on ONNX Runtime (CC BY-NC-SA 4.0, see the folder LICENSE).

Citation

@INPROCEEDINGS{11236178,
  author={Humonen, Innokentiy and Golyadkin, Maksim and Kalin, Danil and Makarov, Ilya},
  booktitle={2025 IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct)}, 
  title={Automatic Image Translation of Long Ancient Egyptian Texts for Augmented Reality Applications}, 
  year={2025},
  volume={},
  number={},
  pages={875-876},
  keywords={Translation;Optical character recognition;Pipelines;Solids;Augmented reality;OCR;image translation;augmented reality},
  doi={10.1109/ISMAR-Adjunct68609.2025.00234}}