Custom attribute image generation based on improved StyleGAN2
Guochao Gao, Xiaoli Li, Zhenlong Du · 2023
The continuous development of generative adversarial networks [1] has made it easier and easier to generate forged face images. Using generative adversarial networks can easily generate a large number of face images. However, the face images generated by generative adversarial networks are not of high quality and the face attributes cannot be controlled. In order to solve these problems, In this paper, we first use a style-based generative adversarial network [2] as a face image generator to generate high-quality face images, and then we create a convolutional neural network [3] that can predict the age and gender attributes of the input face image to control the age and gender of the generated face image. We experimentally demonstrated the reliability of the idea and finally achieved face image generation with customized attributes.