AtlasOCR: Building the First Open-Source Darija OCR Model with Vision Language Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866917395258408960 |
|---|---|
| author | Momayiz, Imane Elaouad, Soufiane Ait Elmajjodi, Abdeljalil Bouanane, Haitame |
| author_facet | Momayiz, Imane Elaouad, Soufiane Ait Elmajjodi, Abdeljalil Bouanane, Haitame |
| contents | Darija, the Moroccan Arabic dialect, is rich in visual content yet lacks specialized Optical Character Recognition (OCR) tools. This paper introduces AtlasOCR, the first open-source Darija OCR model built by fine-tuning a 3B parameter Vision Language Model (VLM). We detail our comprehensive approach, from curating a unique Darija-specific dataset leveraging both synthetic generation with our OCRSmith library and carefully sourced real-world data, to implementing efficient fine-tuning strategies. We utilize QLoRA and Unsloth for parameter-efficient training of Qwen2.5-VL 3B and present comprehensive ablation studies optimizing key hyperparameters. Our evaluation on the newly curated AtlasOCRBench and the established KITAB-Bench demonstrates state-of-the-art performance, challenging larger models and highlighting AtlasOCR's robustness and generalization capabilities for both Darija and standard Arabic OCR tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_08070 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | AtlasOCR: Building the First Open-Source Darija OCR Model with Vision Language Models Momayiz, Imane Elaouad, Soufiane Ait Elmajjodi, Abdeljalil Bouanane, Haitame Computer Vision and Pattern Recognition Artificial Intelligence Darija, the Moroccan Arabic dialect, is rich in visual content yet lacks specialized Optical Character Recognition (OCR) tools. This paper introduces AtlasOCR, the first open-source Darija OCR model built by fine-tuning a 3B parameter Vision Language Model (VLM). We detail our comprehensive approach, from curating a unique Darija-specific dataset leveraging both synthetic generation with our OCRSmith library and carefully sourced real-world data, to implementing efficient fine-tuning strategies. We utilize QLoRA and Unsloth for parameter-efficient training of Qwen2.5-VL 3B and present comprehensive ablation studies optimizing key hyperparameters. Our evaluation on the newly curated AtlasOCRBench and the established KITAB-Bench demonstrates state-of-the-art performance, challenging larger models and highlighting AtlasOCR's robustness and generalization capabilities for both Darija and standard Arabic OCR tasks. |
| title | AtlasOCR: Building the First Open-Source Darija OCR Model with Vision Language Models |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2604.08070 |