aungthuhein-dev/myanmar-handwriting

Learn Myanmar Script

Overview

Learn Myanmar Script is an educational Myanmar handwriting recognition project designed to help beginners practice and recognize Myanmar letters through a browser-based interface. The application allows users to draw Myanmar characters on a canvas, compare their writing to reference forms, and learn through a simple quiz mode.

This project is intended as a lightweight demonstration for teaching Myanmar script and supporting interactive learning.

Model Summary

The project uses a small ONNX model for handwritten character recognition. It is optimized for educational use and interactive demos rather than large-scale production OCR or highly generalized document recognition.

Intended Use

• Practice writing Myanmar consonants and vowels
• Learn Myanmar script through visual examples
• Demonstrate browser-based ML inference with ONNX Runtime Web
• Support educational tools and beginner-friendly self-learning

Out of Scope

• High-accuracy document OCR
• Universal recognition across all handwriting styles
• Production legal or official document processing
• Large-scale deployment without additional validation and retraining

Demo

Live demo: https://aungthuhein2005.github.io/learn-myanmar-script/

Project repository: https://github.com/aungthuhein2005/learn-myanmar-script

Model Details

• Framework: ONNX
• Inference library: ONNX Runtime Web
• Deployment type: browser-based static web app
• Use case: educational handwritten character recognition demo

Data and Training Notes

This model is for demonstration and learning purposes. It is not intended as a fully benchmarked, general-purpose handwriting recognition model. Recognition quality depends on stroke style, spacing, input quality, and the user’s drawing consistency.

Limitations

• Best results are achieved with clean, centered handwriting
• Recognition may vary with different writing styles
• Some Myanmar characters and stroke combinations may be harder to detect
• This is not a production-grade OCR system

Ethical Considerations

This project is intended for education and accessibility. It should not be used as a sole basis for identity verification, official document analysis, or critical decision-making without additional validation and domain-specific review.

Deployment Notes

The web app fetches the ONNX model files from the public repository and runs inference in the browser. This keeps the experience lightweight and avoids requiring a server backend.

Model Usage Example

from huggingface_hub import hf_hub_download

model_path = hf_hub_download(
    repo_id="aungthuhein-dev/myanmar-handwriting",
    filename="model.onnx"  # Replace with the exact filename inside their 'Files and versions' tab
)

print(model_path)

If the model is stored as a split file or has additional artifacts such as model.onnx.data, download them separately using the same repo_id and the exact filename shown in the Hugging Face repository UI.

Citation

@misc{learnmyanmarscript,
  title = {Learn Myanmar Script},
  author = {Aung Thu Hein and Yan Myo Aung},
  year = {2026},
  howpublished = {GitHub and Hugging Face deployment card},
  url = {https://github.com/aungthuhein2005/learn-myanmar-script}
}

License

The frontend code is distributed under the Apache License 2.0. Please review the repository and model files for additional terms if you plan to reuse or extend the project beyond the educational demo.

Authors

• Aung Thu Hein
• Yan Myo Aung

Contact

For questions, feedback, or collaboration, please use the GitHub repository or open an issue in the project.