Archatext/HTR_model_Finetued_Arams-28k

arman-kraken-ocr

Intro

arman-kraken-ocr is a Kraken OCR/HTR model for line-level recognition of handwritten Arabic manuscripts.

The model is based on the Muharaf-trained Kraken checkpoint muharaf_rec_best.mlmodel and was subsequently fine-tuned on the ArMan dataset for historical Arabic manuscript recognition.

• Architecture: Kraken VGSL recognizer
• Kraken version: 5.3.0
• Output classes: 174
• Checkpoint size: ~12 MB
• Base model: Muharaf-trained Kraken model
• Fine-tuning data: ArMan-28k

Details

This checkpoint is a fine-tuned derivative of the Arabic HTR model trained on the Muharaf Corpus:

Bors Uifalean and Simion Toader. HTR Model - Arabic Handwritten Recognition Model Trained on the Muharaf Corpus. Zenodo, 2024. DOI: 10.5281/zenodo.14295489.

The model was fine-tuned on ArMan as part of our work on historical Arabic manuscript HTR and the AraMS-28k dataset.

Citation

If you use AraMS-28k in your research, please cite:

@article{guechaoui2026arams28k,
  title={AraMS-28k: The Largest Publicly Released Line-Level Dataset of Historical Arabic Manuscripts with Margin and Insertion-Anchor Annotations},
  author={Guechaoui, Mohamed and Zellagui, Mohamed Diaa and Chaib, Souleyman and Dhelim, Sahraoui},
  journal={arXiv preprint arXiv:2608.26921},
  year={2026},
  doi={10.48550/arXiv.2608.26921}
}

If you use RefLAM or its annotation pipeline, please also cite:

@article{guechaoui2026reflam,
  title={RefLAM: A Reference-Grounded Line Annotation Pipeline for Historical Arabic Manuscripts},
  author={Guechaoui, Mohamed and Zellagui, Mohamed Diaa and Chaib, Souleyman and Dhelim, Sahraoui},
  journal={arXiv preprint arXiv:2608.25140},
  year={2026},
  doi={10.48550/arXiv.2608.25140}
}

Related papers

• AraMS-28k: https://arxiv.org/abs/2608.26921
• RefLAM: https://arxiv.org/abs/2608.25140

Limitations

• Recognition quality may vary across handwriting styles and manuscript traditions.
• Diacritics may be challenging, particularly when they are absent, faint, or inconsistently represented.
• Image resolution, contrast, noise, cropping, and line segmentation can affect recognition quality.
• The model is intended for handwritten Arabic manuscripts and is not designed for general printed Arabic text.
• Performance on manuscripts outside the fine-tuning domain may differ from performance on ArMan-derived material.