FIGURE 5.

(A) Straw-Net, complete architecture. Here S_i represents different stages of network i ∈ {1,2,…6}. c represents the number of channels inside each SE-ResNet block (we set c = 16), and d is the dilation rate. For details on points P₁ and P₂ see Section “Ablation Study for DAM.” (B) SE-ResNet module; in the first three convolutional blocks, the first value represents the number of input feature maps, the second value represents the kernel size, and the third value represents the number of output channels. The dashed arrows represent the identity mapping. GAP, FC, and r represent global average pooling, densely connected layers, and reduction ratio, respectively.