Learn to abstract via concept graph for weakly-supervised few-shot learning

作者：

Highlights：

• To the best of our knowledge, we are the first to introduce the concept graph and explore the concept hierarchy for addressing the WSFSL problem.

• We propose a novel concept graph-based meta-learning framework, consisting of a multi-level conceptual abstraction-based regularization and a meta concept inference network.

• Extensive experimental results are reported on two realistic datasets, namely, WS-ImageNet-Pure and WS-ImageNet-Mix, which demonstrate the effectiveness of the proposed framework.

摘要

•To the best of our knowledge, we are the first to introduce the concept graph and explore the concept hierarchy for addressing the WSFSL problem.•We propose a novel concept graph-based meta-learning framework, consisting of a multi-level conceptual abstraction-based regularization and a meta concept inference network.•Extensive experimental results are reported on two realistic datasets, namely, WS-ImageNet-Pure and WS-ImageNet-Mix, which demonstrate the effectiveness of the proposed framework.

论文关键词：Few-shot learning,Weakly-supervised learning,Meta-learning,Concept graph

论文评审过程：Received 6 August 2020, Revised 16 December 2020, Accepted 5 March 2021, Available online 22 April 2021, Version of Record 8 May 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107946