Fig. 2.
A schematic of the FD U-net used as the basis for our model. The input matrix size is 256 by 256. The number on the top of each column is the feature maps in each layer. On the right, the yellow block shows the dense connectivity in detail. ReLU: Rectified Linear Unit. BN: Batch normalization. Conv: convolution. The yellow-dashed block shows the dense block at different layers. f1 represents the initial channels of the layer. The growth rate of each layer is four to learn a number of feature maps.
