Abstract—Content-Based Image Retrieval (CBIR) technique attempts to retrieve relevant query images from the extensive repositories of images. With the advancements of the internet and multimedia technology, images have increased at a significant rate. Retrieving similar pictures from a vast database has always been an arduous task where CBIR techniques are helpful. However, similar images retrieval efficiency improvement is a common problem with the available CBIR techniques due to inadequate feature sets. This paper proposes a novel CBIR technique using a Deep Convolutional Neural Network (DCNN)-based AutoEmbedder. With this novel approach, this study attempt to map the higher dimensional features into relevant clusterable embeddings with k-means clustering to cluster the relevant images. The architecture is evaluated using the Corel10K and CIFAR-10 datasets, and the average precision and recall value is used to evaluate the architecture’s performance. The proposed model’s significance is that it outperforms the existing CBIR techniques presented in experimental results.
 

Index Terms—Content-Based Image Retrieval (CBIR), Deep Convolutional Neural Network (DCNN), AutoEmbedder, K-means clustering

Cite: Md. Mohsin Kabir, Adit Ishraq, Kamruddin Nur, and M. F. Mridha, "Content-Based Image Retrieval Using AutoEmbedder," Journal of Advances in Information Technology, Vol. 13, No. 3, pp. 240-248, June 2022.

Copyright © 2022 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.