ne of the perennial challenges of the jewellery industry is the analysis of the multiple categories of gemstones on which the sector is based – its “raw materials”, so to speak. Correctly evaluating the authenticity of a gemstone and its country of origin is a key aspect of jewellery. The dedicated discipline of gemmology is defined as the “science relating to the study of precious, semi-precious, fine and ornamental stones obeying certain aesthetic criteria.”
Given the quantity of gems involved, however, not all of them can be subjected to a thorough and detailed analysis. This has made it necessary to select which ones are worth the effort and cost of testing – that is, until today. Thanks to artificial intelligence and machine learning, the gemmological laboratory of the Swiss jewellery house Gübelin and CSEM, well-known among watchmakers, are developing a tool that will make this pre-selection unnecessary. This technique will make it possible to handle much larger volumes of precious stones.
One of the perennial challenges of the jewellery industry is the analysis of the multiple categories of gemstones on which the sector is based – its “raw materials”, so to speak.
- The Gübelin gemstone reference collection serves as the basis for the daily work of the Gübelin Gem Lab, and consists of more than 27,500 stones. It is probably the world’s most complete collection of rubies, sapphires, emeralds, chrysoberyls, alexandrites and other types of coloured gemstone culled from all the commercially relevant mines worldwide.
“We have been working on gem recognition on a daily basis for almost a century,” says Daniel Nyfeler, Managing Director of the Gübelin Gem Lab. “But with this new technique, we will be able to examine a much more complete set of data. In the past, our clients (i.e gemstone traders) had to select the stones that were deemed significant and the characteristics to be evaluated. But this left some gems out. This new tool democratises the analysis of gemstones, in a way.”
Giving new meaning to the data
Daniel Nyfeler points out: “For example, if a stone had a lower value, it did not make sense to carry out expensive analyses until now. Now we can include all the data in the technology. This will also allow us to take a fresh look at certain gems that were excluded from our analyses.”
The company uses the know-how of CSEM, a specialist in artificial intelligence and machine learning, to automate the gem evaluation process. A new analysis platform will improve the consistency and reliability of data interpretation, reduce potential human errors and save time.
“Gemtelligence”, the software for the automated analysis of gemstones, is based on Gübelin’s extensive catalogue.
- Daniel Nyfeler, Managing Director of the Gübelin Gem Lab
This joint software development project for the automated analysis of gemstones, called “Gemtelligence”, has also been submitted to Innosuisse, a governmental institute whose mission is to promote science-based innovation. The Swiss government has provided “significant funding” for this research.
The research aims in particular to develop machine learning-based algorithms and train them to evaluate standard gemstone charac- teristics. The process is based on an existing catalogue of data from tens of thousands of clients’ gemstones, which the Gübelin Gem Lab has been testing since the 1970s. This data will be complemented by the Gübelin Reference Stone Collection of more than 27,500 gems.
Support from the Swiss government
Philipp Schmid, Head of Industry 4.0 and Machine Learning at CSEM, stresses the heterogeneous nature of the data to be processed: “We are dealing with data in a wide variety of formats, ranging from spectra, chemical element concentrations, microscopy images, handwritten descriptions to subjective evaluations by various experts. The goal is to create a kind of super-expert, working hand-in-hand with the human experts.”
The goal is to create a kind of super-expert, working hand-in-hand with the human experts. The technology should be operational by 2023.
Gübelin and the CSEM have been collaborating for ten years on various projects, but this research takes them into a new dimension, one that has the potential to revolutionise gemmological analysis. “Basically, we want to do the same work as in the past, but in a much more reliable and extensive way,” says Daniel Nyfeler. The Swiss company will offer its expertise in the field to other jewellery manufacturers, who will be able to send their data to its laboratory for processing. The technology should be operational by 2023.
- Philipp Schmid, Head of Industry 4.0 and Machine Learning at CSEM
As Philipp Schmid points out: “The support of Innosuisse and CSEM’s own mission statement mean that we will also be exploring the application of this technology to other industries. The potential is enormous and very diverse, especially for activities that process a large amount of technical data, such as CNC machines, for example.” In the longer run, an application in watchmaking could be a possibility.
GÜBELIN GEM LAB Gübelin operates one of the world’s most respected gemmological laboratories as a completely independent subsidiary. In its laboratories in Lucerne, Hong Kong and New York, the Gübelin Gem Lab provides analyses of diamonds, coloured gemstones, and pearls. Gübelin has also set up its own Academy offering courses in gemmology, as well as the “Provenance Proof” programme aimed at increasing transparency in the gem supply chain through the use of blockchain.
CSEM CSEM, founded in 1984, is a Swiss research and development centre (public-private partnership) specialising in microtechnology, nanotechnology, microelectronics, system engineering, photovoltaics and information and communications technologies. More than 500 specialists from various scientific and technical disciplines work for CSEM in Neuchâtel, Zurich, Muttenz, Alpnach, and Landquart.