Researchers from University Putra Malaysia and UNSW Sydney have built a deep learning system that analyzes and appraises Chinese ceramics — predicting auction values with up to 99% accuracy.
Trained on centuries of decorative and morphological ceramic data, the AI uses a YOLOv11-based model and runs entirely on a consumer-grade NVIDIA GeForce RTX 3090 (not a data center, not cloud). It classifies vessel types, detects design patterns (plants, animals, glaze textures, etc.), and estimates price brackets based on Sotheby’s and Christie’s auction histories.
The goal? Make cultural expertise accessible and scalable — democratizing artifact valuation for collectors, museums, and digital archives without relying solely on rare human experts.
Trained on centuries of decorative and morphological ceramic data, the AI uses a YOLOv11-based model and runs entirely on a consumer-grade NVIDIA GeForce RTX 3090 (not a data center, not cloud). It classifies vessel types, detects design patterns (plants, animals, glaze textures, etc.), and estimates price brackets based on Sotheby’s and Christie’s auction histories.
The goal? Make cultural expertise accessible and scalable — democratizing artifact valuation for collectors, museums, and digital archives without relying solely on rare human experts.