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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Many works have succeeded in reconstructing Gaussian human avatars frommulti-view videos. However, they either struggle to capture pose-dependentappearance details with a single MLP, or rely on a computationally intensiveneural network to reconstruct high-fidelity appearance but with renderingperformance degraded to non-real-time. We propose a novel Gaussian human avatarrepresentation that can reconstruct high-fidelity pose-dependence appearancewith details and meanwhile can be rendered in real time. Our Gaussian avatar isempowered by spatially distributed MLPs which are explicitly located ondifferent positions on human body. The parameters stored in each Gaussian areobtained by interpolating from the outputs of its nearby MLPs based on theirdistances. To avoid undesired smooth Gaussian property changing duringinterpolation, for each Gaussian we define a set of Gaussian offset basis, anda linear combination of basis represents the Gaussian property offsets relativeto the neutral properties. Then we propose to let the MLPs output a set ofcoefficients corresponding to the basis. In this way, although Gaussiancoefficients are derived from interpolation and change smoothly, the Gaussianoffset basis is learned freely without constraints. The smoothly varyingcoefficients combined with freely learned basis can still produce distinctlydifferent Gaussian property offsets, allowing the ability to learnhigh-frequency spatial signals. We further use control points to constrain theGaussians distributed on a surface layer rather than allowing them to beirregularly distributed inside the body, to help the human avatar generalizebetter when animated under novel poses. Compared to the state-of-the-artmethod, our method achieves better appearance quality with finer details whilethe rendering speed is significantly faster under novel views and novel poses.

Real-time High-fidelity Gaussian Human Avatars with Position-based Interpolation of Spatially Distributed MLPs