Grading Green Gems: Can Machines Match Human Eyes?

Color matters a lot when buying or selling gemstones. People often disagree on shades, especially with gems like chrysoprase, whose apple-green tone can look different under various lights. Machines might solve this problem. Researchers measured 51 chrysoprase samples using a spectrophotometer, a device that reads exact colors. They also created 676 reference points for comparison. Using a method called K-means, they grouped the gem colors by similarity. Then, they tested different machine learning models—logistic regression, neural networks, k-nearest neighbors, support vector machines, and random forest—to see which one could accurately sort the gemstones.Logistic regression and neural networks performed best, but the simpler logistic regression model won because it’s easier to understand and faster to run. When tested on real gemstones, the system got every one right. Even when mixing real and fake data, it hit a near-perfect score of 99. 59%. This shows that machines can grade gem colors consistently, something humans struggle with. The team even built a free app where anyone can try the grading system.