Diffusion-Driven GAN Inversion for Multi-Modal Face Image Generation

We present a new multimodal face image generation method that converts a text prompt and a visual input, such as a semantic mask or scribble map, into a photorealistic face image. To do this, we combine the strengths of Generative Adversarial networks (GANs) and diffusion models (DMs) by employing the multimodal features in the DM into the latent space of the pretrained GANs. We present a simple mapping and a style modulation network to link two models and convert meaningful representations in feature maps and attention maps into latent codes. With GAN inversion, the estimated latent codes can be used to generate 2D or 3D-aware facial images. We further present a multi-step training strategy that reflects textual and structural representations into the generated image. Our proposed network produces realistic 2D, multi-view, and stylized face images, which align well with inputs. We validate our method by using pretrained 2D and 3D GANs, and our results outperform existing methods. Our project page is available at https://github.com/1211sh/Diffusion-driven_GAN-Inversion/.

1. Introduction

In recent years, multimodal image generation has achieved remarkable success, driven by the advancements in Generative Adversarial Networks (GANs) [15] and diffusion models (DMs) [11], [18], [48]. Facial image processing has become a popular application for a variety of tasks, including face image generation [21], [39], face editing [6], [12], [30], 3[6], [37], 4[6], and style transfer [7], [64]. Many tasks typically utilize the pretrained StyleGAN [21], [22], which can generate realistic facial images and edit facial attributes by manipulating the latent space using GAN inversion [39], [42], [58]. In these tasks, using multiple modalities as conditions is becoming a popular approach, which improves the user's controllability in generating realistic face images. However, existing GAN Figure 1.

We present a method to map the diffusion features to the latent space of a pretrained GAN, which enables diverse tasks in multimodal face image generation and style transfer. Our method can be applied to 2D and 3D-aware face image generation.

inversion methods [51], [58] have poor alignment with inputs as they neglect the correlation between multimodal inputs. They struggle to map the different modalities into the latent space of the pretrained GAN, such as by mixing the latent codes or optimizing the latent code converted from a given image according to the input text.