Exploring Disentangled Feature Representation Beyond Face Identification
This addresses the need for more comprehensive face representation in computer vision, though it appears incremental as it builds on existing disentanglement methods.
The paper tackles the problem of learning disentangled face features beyond just identity verification, achieving state-of-the-art identity verification on LFW while also enabling competitive face attribute recognition on CelebA and LFWA.
This paper proposes learning disentangled but complementary face features with minimal supervision by face identification. Specifically, we construct an identity Distilling and Dispelling Autoencoder (D2AE) framework that adversarially learns the identity-distilled features for identity verification and the identity-dispelled features to fool the verification system. Thanks to the design of two-stream cues, the learned disentangled features represent not only the identity or attribute but the complete input image. Comprehensive evaluations further demonstrate that the proposed features not only maintain state-of-the-art identity verification performance on LFW, but also acquire competitive discriminative power for face attribute recognition on CelebA and LFWA. Moreover, the proposed system is ready to semantically control the face generation/editing based on various identities and attributes in an unsupervised manner.