Gemological Institute of America and Razor Labs bring advanced AI solution to diamond grading
April 6, 2022
TEL AVIV-Israel – April 6th, Razor Labs (TASE: RZR) and GIA, the Gemological Institute of America, announced today that they have developed a system to apply artificial intelligence (AI) to the crucial task of determining a diamond’s shape and cutting style.
Working closely with GIA, Razor Labs developed an AI algorithm in just six months to power the system that achieved 99% accuracy in predicting diamond shape and cutting style.
GIA’s large, diverse dataset of real-world examples and expert shape determinations from experienced diamond graders provided the ideal foundation for training the Razor Labs’ algorithm. Together they refined the training with detailed shape definitions and mathematical rules to further boost the AI’s accuracy. The system is now being tested in GIA laboratories in diamond centers around the world.
“As GIA introduces more automation into its grading process, this is an important remaining component that has become more challenging to classify,” said Tom Moses, GIA executive vice president and chief laboratory and research officer. “The introduction of more varieties of fancy shapes and cutting styles made this an important initiative to complete.”
Largely unaddressed by automation until now, this critical step offers an excellent opportunity for further application of AI to GIA’s industry-leading diamond grading.
There are more than a dozen standard diamond shapes beyond the instantly recognizable round, including oval, pear, marquise and less frequently seen shapes such as briolette and shield. Adding recognized modifications for those standard shapes brings the number of possible shape determinations to nearly 40.
“GIA is moving ever closer to making diamond grading as automated as possible,” said Pritesh Patel, GIA chief operating officer. “Whether applied to the complexity of determining a diamond’s color and clarity grades or identifying a diamond’s shape, GIA is using AI to dramatically advance diamond grading efficiency and consistency.”
How it works
Michael Zolotov, Razor Labs CTO explained how the AI algorithm works: “To achieve these impressive results, we chose the best neural network architecture for the task and trained it with a carefully prepared dataset drawn from the tens of millions of diamonds examined by GIA. After training and optimizing hyper-parameters, we obtained a combination of networks that can extract and identify the specific visual features that define each diamond’s shape and cutting style and classify them correctly – even in borderline cases.”
Raz Roditti, Razor Labs CEO added: “We remain committed to providing solutions to industrial sectors for more efficient, profitable and sustainable processes. We are very pleased to provide GIA with a disruptive technology that enables higher levels of operational excellence. We look forward to continuing this important collaboration with GIA.”
About GIA
GIA (Gemological Institute of America) is an independent, nonprofit organization established in 1931. Recognized as the world’s foremost authority in gemology, GIA is known for introducing the industry’s 4C standard of Color, Clarity, Cut and Carat Weight. In 1953, they created the International Diamond Grading System™ which is recognized till this day as the standard for diamond quality around the world.
Through research, education, gemological laboratory services and instrument development, the Institute is dedicated to ensuring the public trust in gems and jewelry by upholding the highest standards of integrity, academics, science and professionalism
About Razor Labs
An AI industry leader, Razor Labs is on a mission to transform leading industries into smarter, safer and more efficient operations, helping them unlock greater value and become more sustainable. Razor Labs embodies the spirit of Israeli innovation together with a forward-thinking global perspective, and together with its multinational partners is driving change across multiple industries and numerous fields.
