The invention discloses a portrait generation method based on the
face structure information. The method comprises the following steps: 1, carrying out data preprocessing on an original image, a target image and the
face structure information; 2, carrying out
feature extraction and
feature fusion at the input end of an image generator through a
face structure information model; 3, adopting a combination
loss function based on face structure components for the
loss function part of the image generator; 4, generating a
countermeasure network by using the generator, and distinguishing by using adiscriminator; 5, carrying out model training, and training neural network parameters through a reverse propagation
algorithm. According to the invention, a neural
network model generated from a facephoto to a portrait is provided. The invention particularly provides a method for carrying out
network parameter optimization by utilizing the guidance information of a face part through a portrait generator to generate a portrait and calculating the loss of each part by means of the part information so as to optimize network parameters. Based on the method, the best effect in the generation fieldof face photos and portraits is obtained.