reviseIn the REA-Net we utilize the binary cross-entropy loss to minimize the discrepancy between the predicted output and ground-truth The training process involves a total loss which contains decode

In the REA-Net, we employ the binary cross-entropy loss to minimize the difference between the predicted output and the ground-truth. The training process includes a overall loss, which consists of the decoder loss and the joint loss of the multi-task learning module. It can be defined as follows: