| Popis: |
Background: In recent times, content-based image retrieval (CBIR) remains a hot research topic due to the popularity of Internet and low-cost imaging technologies. In CBIR, the query image (QI) gets matched with the visual details of the images kept in the dataset, and the images with maximum similarity are retrieved. The advent of artificial intelligence (AI) and deep learning (DL) models enables to effective design of the CBIR model. Purpose: In this view, this paper presents an optimal DL based VGG-16 model with grasshopper optimization algorithm (DLVGG-GOA) for CBIR. Methods: The presented model involves the DL based VGG-16 model for the extraction of feature vectors of the applied input query image (QI) and the images in the database. In order to increase the effective performance of the VGG-16 model, the GOA is applied for adjusting the hyperparameters of the VGG-16 model. Upon inputting a QI, the presented model extracts the feature vectors and matches the similarity measurement using Minkowski distance with the images that exist in the database. Finally, the images with higher similarity get retrieved effectively from the database. Results: The retrieval performance of the DLVGG-GOA model is validated using the benchmark database and examines the results interms of distinct measures. Conclusion: The obtained retrieval results ensured the improved effectiveness of the DLVGG-GOA model on the applied test images. |
