《Dual Encoding U-Net for Retinal Vessel Segmentation》阅读笔记-MICCAI2019

做者：Bo Wang1,2, Shuang Qiu2, and Huiguang He1,2,3

目的：Retinal Vessel Segmentation is an essential step for the early diagnosis of eye-related diseases, such as diabetes and hypertension. Segmentation of blood vessels requires both sizeable receptive field and rich spatial information.

方法：Dual Encoding U-Net (DEU-Net), 空间information和上下文information

该结构图outputpatches居然和input同样。

1. Spatial Path，连续四个卷积，用了大stride：7，12，9，6.参考了Large kernel matters-improve semantic segmentation by global convolutional network.， CVPR2017。
2. Context Path，就是inception block，google提出的
3. Attention Skip Module，最简单的attention方式处理
4. Feature Fusion Module，这个方式我看到过，不知道为何叫作feature fusion，其实连结处就是和attention residual for image classification那篇文章同样.
5. Multiscale Predict Module，这个模块没看到过，主要是pixel shuffle（参考CVPR2016 Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network）这个操做。 

 

试验结果：略