Joshua VanderMeer, Michael Vertin, Forrest Hartley, Aidan Hebert

Title:

INSIGHT: AI-Powered Image Search for Cline Library’s SCA

Abstract:

The Cline Library’s Special Collections and Archives (SCA) at Northern Arizona University (NAU) houses a vast digital collection documenting the history of the Colorado Plateau. With over 100,000 digitized items, navigating this extensive archive can be challenging for researchers. This project proposes the development of an image similarity search application that allows users to select an image from their device and retrieve visually similar images from the SCA digital archives. By leveraging modern machine learning techniques, the system will generate image embeddings using vectorization, storing them in a vector database. The search functionality will rely on cosine similarity to determine related images. The application will feature a web-based interface that enables users to upload an image, view results in a structured grid, and interact dynamically with the search results. The architecture consists of a front-end built with React, a Node.js server acting as an API gateway, a vector database powered by Milvus, and machine learning models for embedding generation. The system will utilize TensorFlow and Hugging Face to extract meaningful features from images. This project aims to improve accessibility and search efficiency for SCA users by offering an intuitive and responsive search experience. The implementation will follow a structured development plan, including data processing, UI development, testing, and deployment. The final deliverables will include a fully functional application, comprehensive documentation, and a professionally maintained codebase for future development. Through this innovation, SCA’s digital resources will become more discoverable, enhancing historical research and education.