Bootstrap

医学分割论文:Multi-scale guided attention for medical image segmentation

医学分割论文:Multi-scale guided attention for medical image segmentation

Auther: Ashish Sinha, Institution:Indian Institute of Technology Roorkee, Publish year: 2019
论文地址:https://arxiv.org/abs/1906.02849
github: https://github.com/sinAshish/Multi-Scale-Attention

目前存在的问题:

  1. 多尺度方法(encoder-decoder)会导致多次提取相似的低级特征,造成信息冗余
  2. long-range feature dependencies are not efficiently modeled, resulting in nonoptimal discriminative feature representations associated with each semantic class. 宏观/整体特征没有得到有效建模,导致无法得到与之相关联的语义类(低级别标签)的最优特征表示

针对以上问题,作者提出了:使用引导式的自我注意力机制来获取更丰富的上下文信息来克服这些限制,该方法

;