Context Prior for Scene Segmentation arXiv:2004.01547v1

问题:most approaches rarely distinguish different types ofcontextual dependencies, which may pollute the scene un- derstanding.

以往的上下文关系没有很好的区分类别，对语义造成了污染

解决:创建了Affinity Loss和Context Prior with来区分intra-class and inter- class context 

模型：

 

Affinity Loss：

建立类别之间的关系

To explicitly regularize the network to model the relationship between categories,

L为ground truth L，下采样为与特征图同尺寸的smaller ground truth ~L ，经过one hot encoding变成^L,C为类别数目

损失函数：

 

P为预测的概率图，A为Ideal Affinity Map，Ls为segment loss，λa = 0.4，λu = 1 and λg = 1.