Generalized zero-shot classification via iteratively generating and selecting unseen samples

作者：

Highlights：

• Iteratively generate unseen samples and select unseen samples.

• Make the generated unseen samples similar to the real unseen visual prototypes.

• Select the confident unseen samples by the classifier trained with the generated unseen samples.

• Experimental results show the superiority of the proposed model.

摘要

•Iteratively generate unseen samples and select unseen samples.•Make the generated unseen samples similar to the real unseen visual prototypes.•Select the confident unseen samples by the classifier trained with the generated unseen samples.•Experimental results show the superiority of the proposed model.

论文关键词：Generalized zero shot classification,Generative adversarial network,Selecting confident unseen samples,Generating unseen samples,Unseen visual prototypes

论文评审过程：Received 28 August 2020, Revised 30 November 2020, Accepted 19 December 2020, Available online 24 December 2020, Version of Record 29 December 2020.

论文官网地址：https://doi.org/10.1016/j.image.2020.116115