Machine Learning-Based Gemstone Classification through Analysis of Absorption Band Spectrum Images and Refractive Index

Accurate gemstone identification holds paramount importance in the gemstone industry, traditionally demanding expertise and specialized equipment. This research introduces an advanced computer vision method for automated gemstone classification based on visual attributes. The method utilizes machine learning models trained on a meticulously curated dataset encompassing high resolution gemstone images, annotated absorption band spectrum images, and refractive index values. Empirical validation underscores the efficacy of the approach: Logistic Regression achieves notable accuracy of 92.11% in training and 90.00% in validation for spectrum data, effectively discerning gemstones by their distinctive absorption spectra. For predicting refractive indices, the study employs Random Forest classification, supported by compelling empirical evidence. This approach could revolutionize gemstone classification by automating identification and reducing equipment dependency.