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

Abstract

Existing video retrieval studies support single query. To the best of our knowledge, there is no multi-query video retrieval method. In this study, an efficient and fast multi-query video retrieval method is proposed for queries having different semantics. The metod supports unlimited number of queries. Real valued features representing a video are extracted by a deep network and are converted into binary codes. Database items that simultaneously most closely resemble multiple queries are retrieved by Pareto front method. Efficiency of the method is determined by means of a designed graphical user interface.