Design and Implementation of a Web-Based Museum Artifact Recognition System Using Deep Learning

Abstract

The digital transformation of museums has intensified the demand for tools that let visitors identify exhibits promptly, yet many existing solutions rely on native mobile applications that require installation and platform-specific maintenance. This paper presents the design and implementation of a web-based museum artifact recognition system driven by deep learning. A requirements-led three-tier architecture decouples the responsive frontend, backend services, and recognition engine. A comparative analysis of YOLOv8-cls, ResNet50, and a cloud recognition service is then used to examine trade-offs among recognition quality, response latency, offline availability, and integration cost. Evaluation on a coarse-grained, single-label subset of a museum open-access collection illustrates how these criteria affect engine selection for browser-based deployment. The study provides an extensible implementation route for widening public access to digital cultural-heritage information without requiring visitors to install a dedicated application.