Enhancing Gemstone Classification Accuracy: A Deep Learning Approach Using CNN, VGG-19, and Xception

Gemstones demand has been steadily on the rise in recent years, driven by various factors. Notably, the soaring popularity of colored gemstones in jewelry designs and the pervasive influence of social media platforms significantly contribute to this trend. A key development in the gemstone market is the growing consumer preference for ethically sourced stones. As consumers become more socially aware, they actively seek out gemstones that are mined and processed with a strong commitment to environmental sustainability and ethical practices. Some gemstones may look alike but their properties, price, rarity may vary, so wrong certification or recognition of the gemstone can be a problem for both dealers and customers. To avoid this there are pre built models like CNN to classify images automatically. In this paper traditional CNN, VGG-19, Xception models are implemented and trained on a large gemstone dataset containing 87 classes. The main aim is to increase the accuracy of the models which can become potential source to classify a gemstone in real time. Finally, it was proved that Xception network is best suited for this task with an accuracy about 85.21 percentage. This marks the rise of deep learning models in the gemstones field