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

Neural Radiance Fields (NeRF) have emerged as a potent paradigm forrepresenting scenes and synthesizing photo-realistic images. A main limitationof conventional NeRFs is that they often fail to produce high-qualityrenderings under novel viewpoints that are significantly different from thetraining viewpoints. In this paper, instead of exploiting few-shot imagesynthesis, we study the novel view extrapolation setting that (1) the trainingimages can well describe an object, and (2) there is a notable discrepancybetween the training and test viewpoints' distributions. We present RapNeRF(RAy Priors) as a solution. Our insight is that the inherent appearances of a3D surface's arbitrary visible projections should be consistent. We thuspropose a random ray casting policy that allows training unseen views usingseen views. Furthermore, we show that a ray atlas pre-computed from theobserved rays' viewing directions could further enhance the rendering qualityfor extrapolated views. A main limitation is that RapNeRF would remove thestrong view-dependent effects because it leverages the multi-view consistencyproperty.

Ray Priors through Reprojection: Improving Neural Radiance Fields for Novel View Extrapolation