irregular handwritten Chinese character dataset，IHCCD

This article collects the first set of irregular handwritten Chinese character dataset (IHCCD), which includes 3755 categories and 30 samples for each category. IHCCD is handwritten by different non-standard handwriting artists on A4 printing paper, and uses a scanner as the input device to convert handwritten character samples into digital image samples. In the process of collecting the dataset, these non-standard handwriting practitioners do not need to write completely in the order of standard Chinese character strokes. They can freely adjust the thickness, length, and position of strokes, enlarge or shrink radicals arbitrarily, change the degree of inclination of Chinese characters, distort character shapes, and disrupt spatial structures, thus achieving the goal of avoiding current text recognition engines. The experiment shows that this dataset has laid the foundation for promoting the development of non-standard handwritten Chinese character recognition, and has facilitated further research and development in this field.

本文收录了首个非规范手写汉字数据集（Irregular Handwritten Chinese Character Dataset，IHCCD），该数据集涵盖3755个汉字类别，每类包含30个样本。该数据集的手写样本由不同的非规范手写创作者在A4打印纸上完成，采集过程以扫描仪作为输入设备，将手写汉字样本转换为数字图像样本。在数据集采集环节，这些非规范手写创作者无需严格遵循标准汉字的笔画书写顺序，可自由调整笔画的粗细、长度与位置，任意缩放汉字偏旁部首，改变汉字的倾斜程度，扭曲字形并打乱其空间结构，以此避开当前文本识别引擎的识别。经实验验证，该数据集为推动非规范手写汉字识别领域的发展奠定了基础，并为该领域的后续研究与开发提供了有力支撑。

搜集汇总

背景与挑战

背景概述

IHCCD是一个专门针对不规则手写汉字的数据集，包含3755个类别，每个类别30个样本，总计112,650个样本，通过扫描A4纸上的手写内容采集。其关键特点是书写者可以自由调整笔画、部首和字符结构，以模拟难以被标准识别引擎处理的手写变体，旨在促进非标准手写汉字识别技术的研究与开发。