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 Tools annual

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 monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

We introduce a method for novel view synthesis given only a singlewide-baseline stereo image pair. In this challenging regime, 3D scene pointsare regularly observed only once, requiring prior-based reconstruction of scenegeometry and appearance. We find that existing approaches to novel viewsynthesis from sparse observations fail due to recovering incorrect 3D geometryand due to the high cost of differentiable rendering that precludes theirscaling to large-scale training. We take a step towards resolving theseshortcomings by formulating a multi-view transformer encoder, proposing anefficient, image-space epipolar line sampling scheme to assemble image featuresfor a target ray, and a lightweight cross-attention-based renderer. Ourcontributions enable training of our method on a large-scale real-world datasetof indoor and outdoor scenes. We demonstrate that our method learns powerfulmulti-view geometry priors while reducing the rendering time. We conductextensive comparisons on held-out test scenes across two real-world datasets,significantly outperforming prior work on novel view synthesis from sparseimage observations and achieving multi-view-consistent novel view synthesis.

Learning to Render Novel Views from Wide-Baseline Stereo Pairs