Chinese Calligraphy Images with Dense Annotations

A large-scale dataset of Chinese calligraphy images collected from historical works. It covers over 95 calligraphy masters and 5 major script styles, created by author TSXu and last updated in February 2026. Each image includes annotations for bounding boxes, modern text transcriptions, author attribution, and font style labels.

Use Cases

• Train object detection models to locate calligraphy characters using the provided bounding box annotations.
• Develop style classification models to identify one of the 5 major script styles from the font style labels.
• Build author attribution models by learning from the annotated calligraphy master labels for over 95 artists.
• Create text recognition systems that transcribe historical calligraphy into modern text using the transcription annotations.

Strengths

• Covers over 95 distinct calligraphy masters.
• Includes annotations for 5 major script styles.
• Provides dense annotations including bounding boxes and transcriptions per sample.

Limitations

• Specific sample count, image resolution, and geographic origin of works are unknown.
• Potential class imbalance across the 95+ masters and 5 script styles is not quantified.
• Dataset size in gigabytes and file formats are unspecified.

Provenance

Source

Historical Chinese calligraphy works.

Collection Method

Collected and annotated for the UniCalli project.

Time Range

Spans multiple historical dynasties.

Freshness

Last updated on 2026-02-26.

Geography

null