Learning unseen visual prototypes for zero-shot classification

作者：

Highlights：

• A novel zero-shot learning method is developed to rectify the hubness and domain shift problem.

• The proposed method exploits the class level visual samples to learn the projection function.

• The unseen visual prototypes are modified by the label correlations and their knns.

• The proposed method outperforms existing methods on 5 zero-shot learning datasets.

摘要

•A novel zero-shot learning method is developed to rectify the hubness and domain shift problem.•The proposed method exploits the class level visual samples to learn the projection function.•The unseen visual prototypes are modified by the label correlations and their knns.•The proposed method outperforms existing methods on 5 zero-shot learning datasets.

论文关键词：Zero-shot classification,Unseen visual prototypes,Semantic correlation,Hubness,Domain shift

论文评审过程：Received 2 November 2017, Revised 15 June 2018, Accepted 19 June 2018, Available online 6 July 2018, Version of Record 12 September 2018.

论文官网地址：https://doi.org/10.1016/j.knosys.2018.06.034