Open cross-domain visual search

作者：

Highlights：

• We open visual category search to multiple domains.

• We introduce domain-specific prototype learners to map visual inputs to a common semantic space where search occurs.

• We propose three open scenarios to search from and within multiple domains simultaneously.

• We compare to three well-established sketch-based search tasks in the closed setting and achieve competitive results.

摘要

•We open visual category search to multiple domains.•We introduce domain-specific prototype learners to map visual inputs to a common semantic space where search occurs.•We propose three open scenarios to search from and within multiple domains simultaneously.•We compare to three well-established sketch-based search tasks in the closed setting and achieve competitive results.

论文关键词：41A05,41A10,65D05,65D17

论文评审过程：Received 1 April 2020, Revised 17 July 2020, Accepted 21 July 2020, Available online 29 July 2020, Version of Record 3 August 2020.

论文官网地址：https://doi.org/10.1016/j.cviu.2020.103045