Chanrith123333/khmer_english_ocr_image_line

---
license: cc-by-4.0
task_categories:
- image-to-text
language:
- km
- en
pretty_name: Khmer English OCR image line
size_categories:
- 10M<n<100M
---
# Khmer English OCR image line Dataset 📄

A large-scale synthetic dataset for training OCR models on **Khmer and English** text. This dataset contains **12 million** high-quality synthetic images of text lines.

## 🎯 Dataset Overview

- **Total Images**: 12,138,214
- **Languages**: Khmer, English, and mixed
- **Format**: Image-text pairs
- **Use Case**: OCR model training

## 📋 Data Fields

- **image**: PIL Image of the text line
- **text**: Ground truth text string

## 💾 Usage

### Load with Hugging Face

```python
from datasets import load_dataset

dataset = load_dataset("mrrtmob/km_en_image_line")

# Access an example
example = dataset['train'][0]
image = example['image'] # PIL Image
text = example['text'] # str
```

### Train with Kiri OCR

```bash
kiri-ocr train \
--hf-dataset mrrtmob/km_en_image_line \
--epochs 50 \
--batch-size 32
```

## 🎨 Dataset Features

- Multiple Khmer and English fonts
- Realistic augmentations (noise, blur, rotation)
- Variable text lengths (5-100 characters)

## 📚 Citation

```bibtex
@dataset{khmer_english_ocr_image_line,
author = {mrrtmob},
title = {Khmer English OCR image line Dataset},
year = {2026},
publisher = {Blizzer},
howpublished = {\url{https://huggingface.co/datasets/mrrtmob/khmer_english_ocr_image_line}}
}
```

## ⚖️ License

CC BY 4.0

## 🔗 Related

- **Kiri OCR Library**: [github.com/mrrtmob/kiri-ocr](https://github.com/mrrtmob/kiri-ocr)

## ☕ Support

- [Buy Me a Coffee](https://buymeacoffee.com/tmob)
- [ABA Payway](https://link.payway.com.kh/ABAPAYfd4073965)