AI-Driven Lost and Found Management System for Institutions

Abstract

Loss of personal belongings remains a persistent challenge in schools, campuses, and public institutions, often resulting in frustration and reduced trust in existing manual lost-and-found procedures. Traditional approaches rely on handwritten logs and physical storage, which limits accessibility, delays recovery, and provides no intelligent way to match lost items with those found. This project proposes an intelligent Lost and Found Management System, a web-based platform designed to automate the reporting, matching, and retrieval of lost items. The system enables users to report lost or found items through structured digital forms incorporating descriptions, categories, locations, dates, and uploaded images. A key innovation of the system is the integration of artificial intelligence for multimodal item matching. The project employs OpenAI’s CLIP model, combining image embeddings and text embeddings to compute similarity scores between lost and found items. This approach improves accuracy by simultaneously considering both visual and textual features. The AI pipeline is developed and evaluated using Google Colab with GPU acceleration, applying metrics such as accuracy, precision, recall, and F1-score, to measure model performance. The backend architecture integrates FastAPI for AI inference services and Laravel (PHP) for system logic and user management, while the frontend is implemented using HTML, CSS, JavaScript, Bootstrap, and React to ensure a responsive and user-friendly interface. Additional system capabilities include administrative verification of submitted items, automated match notifications, real-time status updates, and generation of digital claim receipts. The project follows a prototyping methodology, enabling iterative refinement driven by continuous user feedback. Overall, the system provides a scalable, transparent, and intelligent solution that enhances item recovery rates, promotes ethical handling of found items, and strengthens trust within institutional environments. Keywords: Lost and Found System; CLIP Model; Multimodal Matching; Image-Text Similarity; FastAPI; Laravel; Web Application; Prototyping Methodology