README.md

---
dataset_info:
- config_name: cleaned
  features:
  - name: image
    dtype: image
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 287421496
    num_examples: 9065
  download_size: 304293885
  dataset_size: 287421496
- config_name: uncleaned
  features:
  - name: image
    dtype: image
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 2465051123
    num_examples: 162863
  download_size: 2380627327
  dataset_size: 2465051123
configs:
- config_name: cleaned
  data_files:
  - split: train
    path: cleaned/train-*
- config_name: uncleaned
  data_files:
  - split: train
    path: uncleaned/train-*
---

*Please visit the [GitHub repository](https://github.com/chuuhtetnaing/myanmar-language-dataset-collection) for other Myanmar Language datasets.*

# Burmese OCR Dataset (Hugging Face Format)

This is a reformatted version of [alexbeatson/burmese_ocr_data](https://huggingface.co/datasets/alexbeatson/burmese_ocr_data) converted into the native Hugging Face datasets format for easier loading and integration with modern OCR training pipelines.

## Dataset Description

This dataset contains Burmese text images and their corresponding ground truth text extracted from real-life documents, suitable for training Optical Character Recognition (OCR) models.

### Source

The original data was curated from the [Burma Library](https://www.burmalibrary.org/) archive, which collects and preserves government and NGO documents. Documents were processed using Google Document AI to extract text and bounding boxes, then cropped into single-line text images.

## Dataset Structure

The dataset consists of two subsets:

| Subset | Samples | Description |
|--------|---------|-------------|
| `cleaned` | 9,065 | Manually validated by two individuals. Some errors may still be present. |
| `uncleaned` | 162,863 | Raw images from Google Document AI output, without manual review. |

### Features

- `image`: Single-line text image in PNG format (RGB)
- `text`: Corresponding ground truth text

## Usage
```python
from datasets import load_dataset

# Load the cleaned subset
cleaned = load_dataset("chuuhtetnaing/burmese_ocr_dataset_hf", "cleaned")

# Load the uncleaned subset
uncleaned = load_dataset("chuuhtetnaing/burmese_ocr_dataset_hf", "uncleaned")

# Load both subsets
dataset = load_dataset("chuuhtetnaing/burmese_ocr_dataset_hf")

# Access samples
sample = cleaned["train"][0]
print(sample["text"])
sample["image"].show()
```

## Citation

If you use this dataset, please cite the original source:
```bibtex
@misc {alexanderbeatson,
	author       = { {Alexander Beatson} },
	title        = { Burmese OCR data },
	year         = 2024,
	url          = { https://huggingface.co/datasets/alexbeatson/burmese_ocr_data },
	doi          = { 10.57967/hf/3361 },
	publisher    = { Hugging Face },
	note     = {ORCID: 0000-0002-1829-5965} 
}
```

## Acknowledgments

- Original dataset by [alexbeatson](https://huggingface.co/alexbeatson)
- Data sourced from [Burma Library](https://www.burmalibrary.org/)
- Text extraction powered by Google Document AI