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 (NeRFs) increase reconstruction detail for novel viewsynthesis and scene reconstruction, with applications ranging from large staticscenes to dynamic human motion. However, the increased resolution andmodel-free nature of such neural fields come at the cost of high training timesand excessive memory requirements. Recent advances improve the inference timeby using complementary data structures yet these methods are ill-suited fordynamic scenes and often increase memory consumption. Little has been done toreduce the resources required at training time. We propose a method to exploitthe redundancy of NeRF's sample-based computations by partially sharingevaluations across neighboring sample points. Our UNeRF architecture isinspired by the UNet, where spatial resolution is reduced in the middle of thenetwork and information is shared between adjacent samples. Although thischange violates the strict and conscious separation of view-dependentappearance and view-independent density estimation in the NeRF method, we showthat it improves novel view synthesis. We also introduce an alternativesubsampling strategy which shares computation while minimizing any violation ofview invariance. UNeRF is a plug-in module for the original NeRF network. Ourmajor contributions include reduction of the memory footprint, improvedaccuracy, and reduced amortized processing time both during training andinference. With only weak assumptions on locality, we achieve improved resourceutilization on a variety of neural radiance fields tasks. We demonstrateapplications to the novel view synthesis of static scenes as well as dynamichuman shape and motion.

UNeRF: Time and Memory Conscious U-Shaped Network for Training Neural Radiance Fields