GaborMadarasz/synthetic_OCR_dataset_v2

匈牙利合成OCR数据集是一个用于微调匈牙利语文本光学字符识别（OCR）模型的合成生成图像-文本数据集。该数据集特别注重正确识别匈牙利语特有的重音字符（如á、é、í、ó、ö、ő、ú、ü、ű）。真实文本来源于匈牙利维基百科，图像通过PIL和Albumentations程序化渲染生成，并模拟了四种视觉配置文件以覆盖真实世界文档条件：现代（modern）、扫描（scanned）、档案（archiv）和段落（paragraph）。数据集包含约100,000张图像，采用CC BY-SA 4.0许可证，格式为HuggingFace数据集，包含图像特征和JSON对话格式。The Hungarian Synthetic OCR Dataset is a synthetically generated image–text dataset for fine-tuning OCR models on Hungarian text, with a particular focus on correct recognition of Hungarian-specific accented characters (á, é, í, ó, ö, ő, ú, ü, ű). The ground-truth text is sourced from Hungarian Wikipedia. Images are rendered programmatically using PIL and augmented with Albumentations, across four visual profiles that simulate real-world document conditions: modern, scanned, archiv, and paragraph. The dataset contains approximately 100,000 images, is licensed under CC BY-SA 4.0, and is formatted as a HuggingFace dataset with image features and JSON conversation format.