Timka28/cyrillic_small

Cyrillic Handwriting Mixed Dataset (small) 是一个平衡的混合西里尔手写文本数据集的子集，适用于手写识别（HWR）、光学字符识别（OCR）、多模态模型和快速实验。数据集包含37,831个样本，分为训练集（32,299个样本）和测试集（5,532个样本）。每个样本包含图像（datasets.Image / Pillow格式）、文本转录、数据集来源、最终拆分（train/test）以及原始数据集中的拆分（如果存在）。数据集来源包括多个公开数据集，如cyrillic-handwriting-dataset、handwritten_ru_letters、comnist等，并遵循特定的训练/测试拆分规则。文本统计显示最小长度为1个字符，最大长度为3,730个字符，平均长度为87.1个字符，中位数为7个字符。

The Cyrillic Handwriting Mixed Dataset (small) is a balanced subset of combined Cyrillic handwritten text datasets, suitable for HWR/OCR, multimodal models, and quick experiments. It contains 37,831 samples, split into train (32,299 samples) and test (5,532 samples). Each sample includes an image (datasets.Image / Pillow format), text transcription, dataset source, final split (train/test), and the original dataset split (if available). The dataset is compiled from multiple public sources, such as cyrillic-handwriting-dataset, handwritten_ru_letters, comnist, etc., with specific train/test split rules. Text statistics show a minimum length of 1 character, maximum length of 3,730 characters, average length of 87.1 characters, and median of 7 characters.