Meta-learning based relation and representation learning networks for single-image deraining

作者：

Highlights：

• We propose the meta-learning based relation and representation learning networks for single-image deraining.

• Our proposed method aims to learn the transferable embeddings of rainy images by characterizing the relation between rainy/clean images.

• Effectiveness of our proposed method is validated through evaluations on different settings by comparing against several state-of-the-art algorithms.

摘要

•We propose the meta-learning based relation and representation learning networks for single-image deraining.•Our proposed method aims to learn the transferable embeddings of rainy images by characterizing the relation between rainy/clean images.•Effectiveness of our proposed method is validated through evaluations on different settings by comparing against several state-of-the-art algorithms.

论文关键词：Meta-learning,Relation network,Single-image deraining,Representation learning

论文评审过程：Received 24 September 2020, Revised 25 January 2021, Accepted 23 February 2021, Available online 26 June 2021, Version of Record 6 July 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108124