Coarse-to-fine pseudo supervision guided meta-task optimization for few-shot object classification
作者：
Highlights：
• A taxonomy and comprehensive survey of Few-Shot Learning (FSL) is illustrated.
• Unsupervised FSL eliminates the dependency on annotations of source dataset in vanilla FSL.
• Our unsupervised FSL method C2FPS-ML explores coarseto fine pseudo supervisions for optimizing meta-task sampling.
• A comprehensive experimental evaluation of C2FPS-ML is conducted.
• C2FPS-ML is competitive with other state-of-the-art Unsupervised FSL methods.
• C2FPS-ML achieves the higher training efficiency than that of the related methods.
摘要
•A taxonomy and comprehensive survey of Few-Shot Learning (FSL) is illustrated.•Unsupervised FSL eliminates the dependency on annotations of source dataset in vanilla FSL.•Our unsupervised FSL method C2FPS-ML explores coarseto fine pseudo supervisions for optimizing meta-task sampling.•A comprehensive experimental evaluation of C2FPS-ML is conducted.•C2FPS-ML is competitive with other state-of-the-art Unsupervised FSL methods.•C2FPS-ML achieves the higher training efficiency than that of the related methods.
论文关键词：Unsupervised few-shot learning,Meta-learning,Clustering,Object classification
论文评审过程：Received 1 April 2021, Revised 28 August 2021, Accepted 31 August 2021, Available online 2 September 2021, Version of Record 12 September 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.108296
