Unsupervised Disentanglement of Linear-Encoded Facial Semantics
This work addresses the need for unsupervised representation learning in facial analysis, offering a flexible method for applications like data augmentation, though it is incremental as it builds on existing StyleGAN and linear regression techniques.
The paper tackles the problem of disentangling facial semantics from StyleGAN without external supervision, achieving interpretable latent representations that align with human intuition and enabling data augmentation for unbalanced data.
We propose a method to disentangle linear-encoded facial semantics from StyleGAN without external supervision. The method derives from linear regression and sparse representation learning concepts to make the disentangled latent representations easily interpreted as well. We start by coupling StyleGAN with a stabilized 3D deformable facial reconstruction method to decompose single-view GAN generations into multiple semantics. Latent representations are then extracted to capture interpretable facial semantics. In this work, we make it possible to get rid of labels for disentangling meaningful facial semantics. Also, we demonstrate that the guided extrapolation along the disentangled representations can help with data augmentation, which sheds light on handling unbalanced data. Finally, we provide an analysis of our learned localized facial representations and illustrate that the semantic information is encoded, which surprisingly complies with human intuition. The overall unsupervised design brings more flexibility to representation learning in the wild.