BriefGPT.xyz
Jun, 2019
半监督语义分割需要强大和多样化的扰动
Consistency regularization and CutMix for semi-supervised semantic segmentation
HTML
PDF
Geoff French, Timo Aila, Samuli Laine, Michal Mackiewicz, Graham Finlayson
TL;DR
在分析语义分割问题时，我们发现其分布没有展现出低密度区域，并将其作为解释半监督分割为何是具有挑战性的问题的原因，然后找到增强选择作为获得可靠性性能的关键，并发现这些改进后的CutOut和CutMix增强技术在标准数据集中提供了最先进的半监督语义分割结果。
Abstract
consistency regularization
describes a class of approaches that have yielded ground breaking results in
semi-supervised classification
problems. Prior work has established the cluster assumption -- under which th
→