Multi-Query Image Retrieval Based on Deep Learning and Pareto Optimality

Abstract

In this study, a method for fast and efficient multi-query image retrieval from large scale databases is introduced. Images used as queries are semantically different from each other. In order to obtain similarity between multiple queries and each item in the database, image features are extracted from a deep networks and then they are converted into binary codes. The database items that simultaneously most closely resemble multiple queries are obtained by the Pareto front method. Furthermore, the method is tested on a designed graphical user interface.