Researchers at University Putra Malaysia, alongside colleagues at UNSW Sydney, have built an AI system that can classify Chinese ceramics and predict their value using deep learning. The tool analyzes decorative motifs, shapes and kiln-specific craftsmanship, and it predicts price categories based on real auction data from institutions like Sotheby’s and Christie’s.
The system achieved test accuracy as high as 99%. It also uses a YOLOv11-based detection model paired with an algorithm that learned market value from years of real-world auction results.
Siqi Wu, one of the researchers behind the project, said: “Artifact pricing and dating still heavily rely on expert judgment.” The team says its aim is to make cultural appraisal more objective, scalable and accessible to a wider audience, including younger collectors, smaller institutions and digital archive projects.
The AI also identifies ceramic vessel shapes using a typological classification system based on modular morphological parts like the bottle neck, handle, shoulder, spout, body and base. It classifies forms such as bottles, jars, plates, bowls, cups, pots and washbasins.
Beyond form, the system analyzes decorative patterns found on Chinese ceramics, grouped into six major categories: plant patterns, animal motifs, landscapes, human figures, crackled glaze patterns and geometric designs.
In one test, the AI assessed a Ming Dynasty artifact at roughly 30% below its final hammer price. The researchers are also exploring AI for other forms of cultural visual heritage, from Cantonese opera costumes to historical murals.
The project runs on an NVIDIA GeForce RTX 3090, a consumer-grade GPU. The researchers say it shows how a gaming-focused graphics card can be used to process centuries of craftsmanship and support work in cultural heritage analysis.
Source: blogs.nvidia.com.
Companies can share verified announcements through Newz9’s international press release submission page.
