云墨济心：历代书法风格数据集

本数据集由同济大学叶晨老师团队整理，源自以观书法机构的历代名家高清碑帖与墨迹扫描件。数据规模达数十万级，覆盖篆隶楷行草五大书体及百余位名家，包含高分辨率单字二值图像与详细元数据（作者、书体、朝代）。在加工方法上，项目使用“基于视觉特征分类的自适应自动化流水线”。利用SVM技术智能分流碑刻与墨迹，实施自适应反色、二值化及形态学去噪，结合人机协同的筛选细粒度标注，保留了笔墨神韵并去除了石花和噪点.

This dataset was curated by the team led by Professor Ye Chen from Tongji University, sourced from high-resolution scanned stele inscriptions and ink works of renowned calligraphers throughout history provided by Yiguan Calligraphy Institution. The dataset has a scale of hundreds of thousands of samples, covering the five major script styles: seal script, clerical script, regular script, running script and cursive script, as well as works from over 100 famous calligraphers. It includes high-resolution binary images of individual characters and detailed metadata including author, script style and dynasty. In terms of processing methodology, the project adopted an "adaptive automated pipeline based on visual feature classification". Specifically, Support Vector Machine (SVM) technology was used to intelligently distinguish between stele inscriptions and ink works, followed by adaptive color inversion, binarization and morphological denoising. Combined with human-machine collaborative screening and fine-grained annotation, this pipeline preserves the artistic charm of brushstrokes and ink while removing stone stains and image noise.

搜集汇总

背景与挑战

背景概述

云墨济心：历代书法风格数据集是一个大规模、高质量的书法数据集，由同济大学团队构建，覆盖篆、隶、楷、行、草五大书体，收录310余位名家的317,574张单字图片，并经过自适应反色、二值化和去噪处理，确保图像质量。数据集提供细粒度标注，包括作者-书体层级化标签、Unicode编码和朝代信息，适用于书法风格识别、生成式书法大模型和数字人文研究等应用场景。