The first approach to segmentation we'll discuss is called U-Net (U-Net: Convolutional Networks for Biomedical Image Segmentation, https://arxiv.org/abs/1505.04597). The name comes from the visualization of the network architecture. U-Net is a type of fully convolutional network (FCN), called so because it contains only convolutional layers and doesn't have any fully connected layers. An FCN takes the whole image as input, and outputs its segmentation map in a single pass. We can separate an FCN into two virtual components (in reality, this is just a single network):
- The encoder is the first part of the network. It is similar to a regular CNN, without the fully connected layers at the end. The role of the ...
