the loss function of each cycle is defined as the mean square error(MSE) between the attention map output At at the tth cycle and mask M, which is the result of subtracting the background image from the input rainy image. The formula is written as follows,