Multi-Type Ancient Chinese Character Recognition (MTACCR) dataset

​​The Multi-Type Ancient Chinese Character Recognition (MTACCR) dataset​​ is a large-scale resource designed to advance research in ancient Chinese script analysis. It is constructed based on the Table of General Standard Chinese Characters (通用规范汉字表), covering ​​7,874 Chinese characters​​ across three levels (3,500 Level-1, 3,000 Level-2, and 1,605 Level-3) and including standard, traditional, and variant forms. With ​​over 9 million samples​​, the dataset features diverse ancient character images—ranging from original scanned manuscripts to segmented glyphs—collected from calligraphic databases, open-source datasets, and data augmentation. MTACCR significantly surpasses existing datasets in character coverage, typological diversity, and scale, providing a comprehensive benchmark for recognition and historical linguistics studies.