همایش بین المللی گوهرشناسی و کانیهای صنعتی

صفحه اصلی / دومین همایش بین المللی گوهرشناسی و کانیهای صنعتی

Advanced Gemstone Recognition Using Convolutional Neural Networks: A Machine Learning Approach

نویسندگان :

Hamid reza Vatanpour¹ Mahdi Shabani²

1- دانشگاه فرهنیگان 2- دانشگاه فرهنیگان

کلمات کلیدی :

Gemstone،classification،convolutional neural networks،image recognition

چکیده :

This paper presents a novel approach to gemstone classification using convolutional neural networks (CNNs). A dataset consisting of 3,200 gemstone images across 87 classes was utilized, with 2,800 images for training and 400 for testing. The proposed CNN architecture, based on a VGG-like model, was designed to automate the identification process, addressing the limitations of traditional handcrafted feature extraction methods. Data augmentation techniques were employed to enhance model generalization. The model achieved an accuracy of 58%, demonstrating its effectiveness in distinguishing between gemstone classes, though challenges persisted with visually similar stones and class imbalance. Despite these challenges, the CNN-based approach outperformed conventional machine learning models, offering a scalable and efficient solution for gemstone recognition. To further assess the model’s performance, an additional test was performed using images from both personal collections and the internet. The model achieved consistent results, demonstrating its ability to generalize beyond the dataset used for training and validation. These tests reinforce the CNN's potential in real-world applications of gemstone identification. Future work will focus on addressing class imbalances, increasing dataset diversity, and exploring advanced machine learning techniques such as transfer learning to further improve classification performance. The findings of this study lay the foundation for further advancements in the field of automated gemstone identification.