Transfer learning model approach for content-based image retrieval proplem

Abstract

In the paper, the author proposes an image feature extraction method using a convolutional neural network model for the problem of content-based image retrieval (Content-Based Image Retrieval - CBIR). The purpose of the proposal is to reduce the semantic gap between low-level features and high-level features, thereby improving the results of image querying. The DenseNet 121 network model combines transfer learning techniques that implement learning to extract features from a database and using the learned knowledge for query image feature extraction. Experimenting on the Corel dataset, the author's proposed method is compared with recently published works, evaluating its effectiveness and applicability in image retrieval systems.