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

3D-aware image synthesis aims at learning a generative model that can renderphoto-realistic 2D images while capturing decent underlying 3D shapes. Apopular solution is to adopt the generative adversarial network (GAN) andreplace the generator with a 3D renderer, where volume rendering with neuralradiance field (NeRF) is commonly used. Despite the advancement of synthesisquality, existing methods fail to obtain moderate 3D shapes. We argue that,considering the two-player game in the formulation of GANs, only making thegenerator 3D-aware is not enough. In other words, displacing the generativemechanism only offers the capability, but not the guarantee, of producing3D-aware images, because the supervision of the generator primarily comes fromthe discriminator. To address this issue, we propose GeoD through learning ageometry-aware discriminator to improve 3D-aware GANs. Concretely, besidesdifferentiating real and fake samples from the 2D image space, thediscriminator is additionally asked to derive the geometry information from theinputs, which is then applied as the guidance of the generator. Such a simpleyet effective design facilitates learning substantially more accurate 3Dshapes. Extensive experiments on various generator architectures and trainingdatasets verify the superiority of GeoD over state-of-the-art alternatives.Moreover, our approach is registered as a general framework such that a morecapable discriminator (i.e., with a third task of novel view synthesis beyonddomain classification and geometry extraction) can further assist the generatorwith a better multi-view consistency.

Improving 3D-aware Image Synthesis with A Geometry-aware Discriminator