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

Monocular 3D object localization in driving scenes is a crucial task, butchallenging due to its ill-posed nature. Estimating 3D coordinates for eachpixel on the object surface holds great potential as it provides dense 2D-3Dgeometric constraints for the underlying PnP problem. However, high-qualityground truth supervision is not available in driving scenes due to sparsity andvarious artifacts of Lidar data, as well as the practical infeasibility ofcollecting per-instance CAD models. In this work, we present NeurOCS, aframework that uses instance masks and 3D boxes as input to learn 3D objectshapes by means of differentiable rendering, which further serves assupervision for learning dense object coordinates. Our approach rests oninsights in learning a category-level shape prior directly from real drivingscenes, while properly handling single-view ambiguities. Furthermore, we studyand make critical design choices to learn object coordinates more effectivelyfrom an object-centric view. Altogether, our framework leads to newstate-of-the-art in monocular 3D localization that ranks 1st on theKITTI-Object benchmark among published monocular methods.

NeurOCS: Neural NOCS Supervision for Monocular 3D Object Localization