CodePudding user response:
1, access to the same category and the target image of a pair of random images, lead the enhancement of network map them by CNN for synthetic images, synthetic image and target image contrast calculation loss; 2, the synthetic image and target image converted into nerve style classification in the network, and output the image classification loss; 3, will strengthen and classification, after the loss of a weighted average of the back propagation network to update the classification and strengthen the network weight, make the output image within the same gap decreases and accurate classification,