Table 2 The specific procedure of the FFM

Input: input_1, input_2

Output: x

1. Concatenate input_1 and input_2 along the channel dimension to create a new tensor x

2. Apply ConvBlock operation to x, yielding a feature tensor

3. Utilize the AdaptiveAvgPool2d operation on the feature tensor, resulting in x

4. Apply Conv2d with ReLU and Sigmoid activations to x

x = Sigmoid(Conv2d(ReLU(Conv2d(AdaptiveAvgPool2d(ConvBlock([input_1,input_2])),kernel_size = 1)), kernel_size = 1))

5. Perform element-wise multiplication of the result of ConvBlock([input_1, input_2]) by x

6. Conduct element-wise addition of the above result to ConvBlock([input_1, input_2])

7. Return x