Finding Task-Relevant Features for Few-Shot Learning by Category Traversal論文筆記

Finding Task-Relevant Features for Few-Shot Learning by Category Traversal 原文爲CVPR 2019 Oral 原文鏈接 Motivation： 文章指出很多Few shot的工作只關注query和support之間的特徵的區別（inter-class），而忽略了support set集合裏面本身的特徵就有區別（intra-

>>阅读原文<<