Stepwise Synthetic Face Sketch Generation Algorithm Based on GAN

Face sketch has rich shadow, texture and vivid facial features, and is widely used in face recognition and life entertainment. In view of the sequence of the drawing steps of the artist’s face sketch, an algorithm is proposed to simulate the steps of the artist’s face sketch. The process of sketch generation is divided into two stages: the first stage uses the reconstructed image to generate sketch information; In the second stage, the sketch information obtained in the first stage is used to assist the synthesis of the target sketch image to achieve the purpose of having the face outline, hair texture, facial features and facial shadow features. CUHK and CUFS data sets were used to conduct a large number of experiments, and three objective evaluation indicators, PSNR, SSIM and FIR, were compared. The results showed that PSNR of the proposed algorithm increased by 4.7 dB on average, SSIM increased by 0.08 on average, and FID score decreased by 8.3 on average compared with typical algorithms; This algorithm can generate better face sketch images.