DatarrX/myanmar-synthetic-syllable-glyphs

Myanmar Synthetic Syllable Glyphs (MSSG) 是一个大规模、高保真的合成图像数据集，包含14,295,552个经过重度增强的字形图像（128x64像素，灰度图），代表了缅甸文字的结构组合矩阵。该数据集由Khant Sint Heinn开发，由DatarrX（缅甸开源组织，非营利组织）正式发布和维护，旨在推进缅甸语的低资源自然语言处理（NLP）、光学字符识别（OCR）和手写文本识别（HTR）框架。数据集通过生成性组合网格方法，自动映射缅甸语的“အ Pattern”规则，并系统替换36个主要字素（包括33个传统缅甸辅音以及ဉ、ဥ和ဧ），从而捕捉缅甸语正字法的详尽演化蓝图。此外，数据集包含非标准、罕见和历史未使用的音节结构，以增强视觉模型的结构鲁棒性和历史计算完整性。图像为8位灰度，具有高对比度的白色前景笔画和纯黑色背景，并应用了多种数据增强技术，如几何旋转、像素偏移和高斯模糊模拟。

The Myanmar Synthetic Syllable Glyphs (MSSG) is a massive-scale, high-fidelity synthetic image dataset containing 14,295,552 heavily augmented glyph images (128x64 pixels, grayscale) representing the structural combinatorial matrix of the Burmese script. Developed and engineered by Khant Sint Heinn, this core foundational dataset is officially published and maintained under DatarrX (Myanmar Open Source Organization, NPO) to advance low-resource Natural Language Processing (NLP), Optical Character Recognition (OCR), and Handwritten Text Recognition (HTR) frameworks for the Myanmar language. The dataset treats the Myanmar language as a generative combinatorial grid, creating an automated pipeline that maps out the extensive အ Pattern rules and systematically substitutes the 36 primary graphemes (the 33 traditional Burmese consonants plus ဉ, ဥ, and ဧ), capturing an exhaustive evolutionary blueprint of native Burmese orthography. It includes non-standard, rare, and historically unutilized syllable structures to enhance structural robustness and historical computational completeness. Images are in 8-bit grayscale with high-contrast white foreground strokes on a pure black background, and data augmentations such as geometric rotations, pixel-shifting, and Gaussian blur simulations are applied.