Makeup migration method based on adaptive instance normalization

The invention discloses a makeup migration method based on adaptive normalization. The method comprises the following steps: step 1, constructing a data set; 2, extracting a makeup feature map and a makeup feature map; step 3, constructing a training network model of makeup migration based on adaptive normalization; 4, inputting the makeup feature image set and the makeup feature image set into a makeup network G1; 5, the makeup network generates an image 1; step 6, inputting the generated image 1 and the corresponding makeup feature map into a discriminator D1, and outputting a discrimination value; 7, inputting the generated image 1 into a makeup removal network G2 to obtain a generated image 2; 8, inputting the generated image 2 and the tone-up feature map into a discriminator D2, and outputting a discriminant value; step 9, returning to the step 5 until any one of the makeup feature image set and the makeup feature image set is empty; and step 10, obtaining trained parameters when a preset number of itera