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

The advent of generative radiance fields has significantly promoted thedevelopment of 3D-aware image synthesis. The cumulative rendering process inradiance fields makes training these generative models much easier sincegradients are distributed over the entire volume, but leads to diffused objectsurfaces. In the meantime, compared to radiance fields occupancyrepresentations could inherently ensure deterministic surfaces. However, if wedirectly apply occupancy representations to generative models, during trainingthey will only receive sparse gradients located on object surfaces andeventually suffer from the convergence problem. In this paper, we proposeGenerative Occupancy Fields (GOF), a novel model based on generative radiancefields that can learn compact object surfaces without impeding its trainingconvergence. The key insight of GOF is a dedicated transition from thecumulative rendering in radiance fields to rendering with only the surfacepoints as the learned surface gets more and more accurate. In this way, GOFcombines the merits of two representations in a unified framework. In practice,the training-time transition of start from radiance fields and march tooccupancy representations is achieved in GOF by gradually shrinking thesampling region in its rendering process from the entire volume to a minimalneighboring region around the surface. Through comprehensive experiments onmultiple datasets, we demonstrate that GOF can synthesize high-quality imageswith 3D consistency and simultaneously learn compact and smooth objectsurfaces. Code, models, and demo videos are available athttps://sheldontsui.github.io/projects/GOF

Generative Occupancy Fields for 3D Surface-Aware Image Synthesis