Rib Fracture Detection with Dual-Attention Enhanced U-Net

Figure 4

(a) The proposed channel-wise fuse attention module (CFAM). denotes the feature map from the encoder path, and denotes the feature map from the decoder path. GAP represents the global average pooling. The symbol next to the feature indicates the size of the corresponding feature. (b) The proposed spatial-wise group attention module (SGAM). denotes the feature map from the decoder block, and the symbol in each block represents the number of channels of each block.

(a)

(b)