GEOSTONE DETECTION AI – INTELLIGENT ROCK AND GEMSTONE RECOGNITION SYSTEM

Abstract: GeoStone Detection AI is an intelligent AI-powered geological analysis system developed for the automatic identification and scientific analysis of rocks, minerals, and gemstones using image processing, computer vision, and multimodal Artificial Intelligence technologies. Traditional geological identification methods mainly depend on manual observation, laboratory testing, and expert geological analysis, which are time-consuming, expensive, and inaccessible for many users. To overcome these limitations, the proposed system integrates modern web technologies, cloud computing, and Vision Language Models into a unified full-stack web application capable of performing real-time geological analysis from uploaded images. The system is developed using React.js for frontend implementation and Node.js with Express.js for backend processing, while the core AI engine utilizes the Groq Cloud API integrated with the Llama 4 Scout Vision model for multimodal image understanding and scientific report generation. Uploaded geological images undergo preprocessing operations such as resizing, normalization, noise reduction, and Base64 conversion before AI analysis. The system extracts visual characteristics including texture, color, transparency, crystal structure, and mineral patterns to identify geological specimens accurately and generate detailed geological reports containing physical properties, chemical composition, geological formation, rarity level, industrial applications, commercial value, and safety information. The application is deployed using Vercel and Render.com to provide scalable cloud-based accessibility across desktop and mobile devices without requiring specialized hardware or software installation. Experimental evaluation demonstrates that the proposed system provides fast response time, reliable prediction performance, efficient preprocessing, and user-friendly interaction for real-time geological analysis. The project highlights the practical implementation of multimodal Artificial Intelligence in scientific applications and demonstrates how Vision Language Models can improve accessibility, automation, and efficiency in geological identification systems while providing a scalable foundation for future enhancements such as multilingual support, offline AI processing, GPS-based geological mapping, and advanced 3D mineral visualization.

Keywords: Artificial Intelligence, Computer Vision, Geological Analysis, Rock Classification, Gemstone Identification, Vision Language Model, Deep Learning, Multimodal AI, Image Processing, Cloud Computing, Scientific Report Generation, Full-Stack Web Application.