Porolis Technologies, an AI tech start-up focused in the gemstone industry, and AI Singapore (AISG), a national AI programme launched by the National Research Foundation, Singapore, have successfully collaborated to develop an AI-powered application for gemstone colour grading. This is made possible through AISG’s Short Industry Project (SIP) programme.

The colour of a gemstone is one of its most defining properties, significantly influencing its aesthetic appeal and market value. Whether it is the deep royal-blue of a sapphire or the vibrant pigeon-blood red of a ruby, precise colour grading is essential to determine gemstone quality and ensuring fair valuation in the market.

The cutting-edge solution addresses a major industry problem where the colour grading practice today rely solely on subjective human visual perception and judgment, even amongst the leading global gem labs. Further, there lacks any harmonized industry standards or device for measuring gemstone colour. This leads to inconsistencies and discrepancies in assessment outcomes. When deployed, the AI-powered gemstone colour grading system will not only augment the tasks that is regularly performed by trained gemmologist in gem labs, but also empower key industry stakeholders such as miners, traders, retailers and end-consumers with a convenient tool to objectively and uniformly determine gemstone colour.

Porolis, together with its sister gem lab company, Gemmological Institute of Colombo, contributed a large dataset of approximately 50,000 gemstones samples for training the AI model. The dataset comprises of high-resolution images of sapphires and rubies, representing the full spectrum of gemstones colours that give these precious gemstones their allure. The images have been expertly labelled by Gem-A FGA-certified gemmologists to ensure that the AI model is trained and validated using reliable and accurate data as ground truth.

AI Singapore brought its expertise in AI and machine-learning, and together with Porolis’ in-house AI-Engineers, successfully trained an AI-model that is capable of grading sapphire and ruby gemstone colour, matching the accuracy of expert gemmologists. With sufficient training data, the model is able to analyse every pixel of the high-res image and thus is capable of capturing subtle colour variations that humans may overlook, making the system more consistent and accessible than traditional grading methods. Besides aiming to attain expert-level accuracy with its inferences, a key priority for the AI-Engineers is to develop an AI-system that is highly explainable and not perceived as a black-box that gives no reasoning for its colour grading. This emphasis is critical to help industry stakeholders gain confidence and accept the results of an AI-system that they can understand, by exposing the dataset it is trained on as well as the algorithms used.

With the successful completion of this AI project, Porolis is poised to commercialize its potential and build the world’s first colourimeter for gemstones. It will first focus on sapphires and rubies, owing to its rich spectrum of colours and the importance of accurate colour assessment to determine its commercial value. To impose greater consistency when acquiring gemstone images, Porolis has also invented an advanced optical imaging device, the KROMA™ , that is capable of acquiring colour faithful images of sapphires, rubies and other gemstones. The device is connected to a cloud-based AI-application that analyses the gemstone image, assess its colour and returns an accurate colour grade to the User.

Porolis AI Research Lab (PaiR) multi-disciplinary team of AI-, Optical- and Software- Engineers are already embarking on the next phase of AI-application in the domain of gemmology. The custom-design lens-camera inside KROMA™ is capable of acquiring microscopic images of gemstone inclusions that has formed within the crystal during its natural formation millions of years ago. The presence of certain inclusion type provides strong evidence of the geographical origin and also unnatural treatment applied on the gemstone crystal, and these attributes significantly impact its valuation. Porolis’ object-detection AI model is capable of detecting and categorizing such inclusions, hence providing Users a means to make evidenced-based commercial decisions.

The images of inclusions captured by KROMA™ offer another compelling application that is highly scalable. As the inclusions formed in every gemstone is different, Porolis is harnessing the power of AI to map and generate digital fingerprints to uniquely tag each gemstone based on the inclusion natural patterns. This cutting-edge solution will finally open the door for digital tokenization of high-value gemstones assets and leverage blockchain to build efficiencies and transparency in the trade.

“Our team of engineers and gemmologists are extremely proud of the results and passionate over how AI can bring impactful change to the global gemstone industry,” said Vincent Wong, CEO of Porolis. “Together with KROMA™ , we offer to companies, end-consumers and enthusiasts a cloud-based AI-as-a-Service (AIaaS) solution. We aim to democratize the science of gemstone identification and authentication, by combining it with advanced technologies of photomicrography, AI and blockchain. These AI-driven solutions, when adopted as industry standards to assess gemstone quality, represent the core components of Porolis’ strategic roadmap to build a blockchain-powered gemstone trading ecosystem that is efficient, transparent and trustful.”