24,495 line images of Arabic handwriting paired with ground-truth text transcriptions. The collection features manuscripts spanning the 19th to 21st centuries, providing a longitudinal view of Arabic script evolution.
Use Cases
- Train Arabic handwriting recognition models using the line images and corresponding text transcriptions
- Analyze stylistic changes in Arabic script over time using the 19th-21st century manuscript metadata
- Benchmark line-level OCR performance on historical Arabic scripts using the provided ground-truth text
Strengths
- 24,495 line images of Arabic handwriting
- Ground-truth text transcriptions for every image
- Manuscripts sourced from the 19th, 20th, and 21st centuries