META-MODEL FOR LOW-CODE CONFIGURATION AND DEPLOYMENT OF CONTENT-BASED IMAGE RETRIEVAL SYSTEMS

The object of this study is content-based image retrieval (CBIR) systems with configurable architectures, and the subject is the meta-model for low-code configuration and deployment of CBIR systems. The goal of this work is to develop a meta-model for CBIR systems that enables their low-code configuration and deployment. The proposed meta-model allows users to define a CBIR system by selecting and combining the CBIR components from a predefined catalog, after which deployment artifacts are automatically generated and can be easily deployed. The proposed meta-model formalizes CBIR components: image repository, feature extractor, feature database (logical and physical levels), similarity measure, result aggregator, and user interaction layer; and extends them with two meta-level components: a configuration manager and a deployment engine. The architecture was implemented using Docker for containerization, Spring Boot starters for modularity, and a web-based graphical interface for configuration. A prototype was developed and tested on a dataset of 100 000 images, with systematic variation of component combinations. Experiments confirmed that the meta-model enables rapid reconfiguration and deployment of CBIR systems, allowing the evaluation of performance under different configurations. The examined difference between the best and worst tested configurations highlighting the significant effect of component selection on system performance. Scientific novelty lies in introducing a formalized meta-model that integrates low-code principles into CBIR design, combining modular architecture, containerized deployment, and graphical configuration in a single framework. The practical significance of the solution is in simplifying CBIR experimentation for researchers and practitioners without deep programming expertise, enabling rapid prototyping, testing, and deployment of customized CBIR systems.