| --- |
| configs: |
| - config_name: default |
| data_files: |
| - split: train |
| path: data/train-* |
| - split: validation |
| path: data/validation-* |
| - split: test |
| path: data/test-* |
| --- |
| |
| # Arabic-Image2Html Dataset |
|
|
| A dataset of **28K image-HTML pairs** for training OCR models to transform Arabic documents into structured HTML. |
|
|
| ## Dataset Description |
|
|
| This dataset contains Arabic document images paired with semantic HTML output. It includes diverse document types and layouts for training image-to-HTML models. |
|
|
| ## Usage |
|
|
| ```python |
| from datasets import load_dataset |
| |
| dataset = load_dataset("abdouuu/arabic-image2html") |
| ``` |
|
|
| ## Dataset Stats |
|
|
| - **Total samples:** 28,044 |
| - **Train:** 22,435 examples |
| - **Validation:** 2,804 examples |
| - **Test:** 2,805 examples |
| - **Language:** Arabic |
| - **Output format:** Semantic HTML |
|
|
| ## License |
|
|
| Apache 2.0 |
|
|
