부산대학교 도서관

In this paper, a reliable gemstone identification system is shown that combines Random Forest as well as CNNs (Convolutional Neural Networks). With a 70:30 training to testing ratio, a diversified dataset of 6265 photos was carefully picked to provide thorough model assessment. The many convolutional and pooling layers added to the CNN architecture effectively identified key information from gemstone photos. This hybrid strategy improved the classification process by integrating a Random Forest classifier. The model's outstanding accuracy of 74.76% demonstrates how adept it is at identifying gemstones. This work shows the possible uses and accountability in more general domains like mineralogy, capacity building, and geological analysis while also advancing automated gemstone categorization. This research area provides a promising precedent for future development in the study of geological specimens by demonstrating the effectiveness of integrating deep learning and machine learning approaches for exact gemstone detection.