StyleGAN-Human: A Data-Centric Odyssey of Human Generation

TL;DR

This paper emphasizes the importance of data engineering in human image generation, introducing a large dataset and analyzing how data size, distribution, and alignment impact StyleGAN's performance.

Contribution

It provides a comprehensive data-centric study with a new large-scale human image dataset and insights into data factors affecting StyleGAN-based human generation.

Findings

Over 40K images are needed for high-quality generation.

Balanced pose data improves quality for rare poses.

Alignment using body centers outperforms face or pelvis anchors.

Abstract

Unconditional human image generation is an important task in vision and graphics, which enables various applications in the creative industry. Existing studies in this field mainly focus on "network engineering" such as designing new components and objective functions. This work takes a data-centric perspective and investigates multiple critical aspects in "data engineering", which we believe would complement the current practice. To facilitate a comprehensive study, we collect and annotate a large-scale human image dataset with over 230K samples capturing diverse poses and textures. Equipped with this large dataset, we rigorously investigate three essential factors in data engineering for StyleGAN-based human generation, namely data size, data distribution, and data alignment. Extensive experiments reveal several valuable observations w.r.t. these aspects: 1) Large-scale data, more than 40K images, are needed to train a high-fidelity unconditional human generation model with vanilla StyleGAN. 2) A balanced training set helps improve the generation quality with rare face poses compared to the long-tailed counterpart, whereas simply balancing the clothing texture distribution does not effectively bring an improvement. 3) Human GAN models with body centers for alignment outperform models trained using face centers or pelvis points as alignment anchors. In addition, a model zoo and human editing applications are demonstrated to facilitate future research in the community.