Cross Modal Image and Text Retrieval Algorithm Based on Deep Learning

Authors

  • Jiana Bi School of Software and Big Data, Changzhou College of Information Technology, Changzhou213164, Jiangsu, China

Abstract

Multi-modal data contains rich and complementary information that provides a more comprehensive means of understanding and analyzing things deeply. Cross-modal image retrieval is related to the effective association between two different modal data, which can accurately retrievea semantically matched image through the input text, or accurately find the corresponding text through the input image. In this study, the algorithm model is built based on deep learning, and the dual tower structure and multi-modal fusion structure are combined to give full play to the advantages of different structures. Image construction based on CNN is responsible for feature extraction of input images, and text construction based on Transformer is used to accurately capture semantic dependencies in text. The addition of a fusion layer to the twin tower structure can better mine the potential
semantic associations between images and text. Two open data sets, MSCOCO and Flickr30k, were used to carry out experiments that demonstrated the superiority of the proposed model in cross-modal image retrieval task.

Keywords: deep learning; cross modal; image and text retrieval; twin tower structure; CNN; Transformer

Cite As

J. Bi, "Cross Modal Image and Text Retrieval Algorithm Based on Deep Learning", Engineering Intelligent
Systems, vol. 34 no. 4, pp. 493-499, 2026.

Downloads

Published

2026-07-01

Issue

Vol. 34 No. 4 (2026): International Journal of Engineering Intelligent Systems

Section

Article

License

The submission of a paper implies that, if accepted for publication, it will not be published elsewhere in the same form, in any language, without the prior consent of the publisher. Before publication, authors are requested to assign copyright to CRL Publishing Ltd. This allows CRL to sanction photocopying, and to authorize the reprinting of issues or volumes according to demand. Authors' traditional rights will not be jeopardized by assigning Copyright in this way, as they retain the right to reuse the material following publication, and to veto third-party publication.