DatarrX/pyu-handwritten-consonant-dataset

Myanmar’s Ancient Heritage: Pyu Handwritten Consonant Dataset

An open-access, systematically curated handwritten dataset of the 33 ancient Pyu consonants. This project serves as a foundational baseline benchmark to support digital humanities, paleographical preservation, and advanced computer vision tasks such as Optical Character Recognition (OCR). The dataset is modeled directly after canonical historical references documented by Thiripyanchi U Tha Myat.

📊 Dataset Specifications

• Total Classes: 33 unique characters (the full canonical Pyu consonant alphabet).
• Original Samples: 825 unique handwritten images (33 consonants × 25 distinct stylistic variations).
• Augmented Samples: 4,950 total images (generated via spatial rotations, translations, Gaussian blur, affine transforms, and perspective distortions).
• Total Images: 5,775 grayscale images.
• Image Specifications: 224 × 224 pixel resolution, single-channel grayscale (white strokes on a solid black background).

📂 Data Structure & Format

This dataset is distributed exclusively in Apache Parquet format for direct, high-performance integration into modern machine learning pipelines (PyTorch, TensorFlow, Hugging Face Datasets).

Images are embedded directly within the file as compressed binary bytes, preserving native characteristics while allowing optimized parallel row access.

Schema Documentation

Each row in the Parquet file contains the following features:

Feature Name	Data Type	Description
image	struct / bytes	Embedded image object (Grayscale, 224x224 PNG format encoded as bytes)
label_1	string	Romanized/English transliteration equivalent (e.g., a)
label_2	string	Modern Myanmar script typographical equivalent (e.g., အ)
note	string	Specification label denoting whether the asset is original or augmented

🚀 Quick Start (How to Use)

You can load this dataset instantly using the Hugging Face datasets library:

from datasets import load_dataset

# Load the dataset directly from Hugging Face Hub
dataset = load_dataset("DatarrX/pyu-handwritten-consonant-dataset")

# Access a single sample
sample = dataset['train'][0]
print(sample['label_1'], sample['label_2'])

# The image is automatically decoded and ready to display
# sample['image'].show()

⚖️ License & Attribution

This dataset is openly distributed under the Creative Commons Attribution 4.0 International (CC BY 4.0) License. You are free to share, copy, and modify the material for any purpose—including commercial applications—provided you give appropriate credit to the original creator.

• Lead Creator & Curator: Khant Sint Heinn (Kalix Louis)
• Organization / Publisher: DatarrX
• Project Intent: Baseline Open-Science Release for Cultural Informatics Preservation