English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Although Generative Adversarial Networks (GANs) have made significantprogress in face synthesis, there lacks enough understanding of what GANs havelearned in the latent representation to map a random code to a photo-realisticimage. In this work, we propose a framework called InterFaceGAN to interpretthe disentangled face representation learned by the state-of-the-art GAN modelsand study the properties of the facial semantics encoded in the latent space.We first find that GANs learn various semantics in some linear subspaces of thelatent space. After identifying these subspaces, we can realisticallymanipulate the corresponding facial attributes without retraining the model. Wethen conduct a detailed study on the correlation between different semanticsand manage to better disentangle them via subspace projection, resulting inmore precise control of the attribute manipulation. Besides manipulating thegender, age, expression, and presence of eyeglasses, we can even alter the facepose and fix the artifacts accidentally made by GANs. Furthermore, we performan in-depth face identity analysis and a layer-wise analysis to evaluate theediting results quantitatively. Finally, we apply our approach to real faceediting by employing GAN inversion approaches and explicitly trainingfeed-forward models based on the synthetic data established by InterFaceGAN.Extensive experimental results suggest that learning to synthesize facesspontaneously brings a disentangled and controllable face representation.

InterFaceGAN: Interpreting the Disentangled Face Representation Learned by GANs