Multistrengthening Module-Based Salient Object Detection

Figure 7

F: unit of the semantic reinforcement unit. The convolution layer other than E₁ is enlarged to the same size as E₁ and finally fused with E₁ to obtain Y_i.