The purpose of the attention discri

The purpose of the attention discrimination network is to distinguish between real images and generated images. Introducing the attention map into the discriminant network, the discriminant network will not only discriminate the global consistency between the raindrop image and the clean image, but also focus on the local consistency of the raindrop region, that is, the attention map would be injected into the discriminative network, assess the image quality and domain difference with the emphasis on the rain area via a local calculation. The discriminative network is based on the one in condition GAN method(The detail is illustrated in Section 3.4.6.4 ), with the exception of the introduction of local assessment.