【Few-Shot Segmentation论文阅读笔记】Prototype Mixture Models for Few-Shot Semantic Segmentation, ECCV, 2020
AbstractTargetProbelm:SinglePrototype===>semanticambiguityproblem为此,本文提出了PrototypeMixtureModels(PMMs)PMMs使用多个prototype用来分别对应不同的imageregions,进而提高其语义表征能力使用EM算法来估算prototype,使得PMMs能够富含丰富的channel-wised和spa