| | --- |
| | 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 |
| |
|
