Pottery as a mixture of earth, fire and art has been part of the global cultural dialogue for thousands of years. Now, researchers in a lab at University of Putra Malaysia and University of New South Wales in Sydney, have linked this millennia-old heritage to the world of silicon.
This team has developed an AI system that can classify Chinese ceramics with stunning accuracy and predict their monetary value. The tool, which uses deep learning, analyzes decorative patterns, exterior shape, and construction details unique to different kilns. The interesting point is that this system, based on actual data from major auctions like Sotheby's and Christie's, estimates the price category of works with 99 percent accuracy.
The driving force behind this project, contrary to what one might expect, is not a massive data center or expensive industrial hardware, but the GeForce RTX 3090 graphics card; the same popular chip that gamers use to run games like Cyberpunk 2077. The system, which uses the YOLOv7 detection model, has directly learned the market value from the results of several years of auctions.
The main goal of this project is to democratize expertise in the field of archaeology. Si Ki Woo, one of the researchers on the project, says that pricing and dating still heavily rely on expert judgment, a expertise that is not always available to young collectors or small museums. This AI aims to make cultural assessment more objective and accessible.