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

Humans perceive and construct the surrounding world as an arrangement ofsimple parametric models. In particular, man-made environments commonly consistof volumetric primitives such as cuboids or cylinders. Inferring theseprimitives is an important step to attain high-level, abstract scenedescriptions. Previous approaches directly estimate shape parameters from a 2Dor 3D input, and are only able to reproduce simple objects, yet unable toaccurately parse more complex 3D scenes. In contrast, we propose a robustestimator for primitive fitting, which can meaningfully abstract real-worldenvironments using cuboids. A RANSAC estimator guided by a neural network fitsthese primitives to 3D features, such as a depth map. We condition the networkon previously detected parts of the scene, thus parsing it one-by-one. Toobtain 3D features from a single RGB image, we additionally optimise a featureextraction CNN in an end-to-end manner. However, naively minimisingpoint-to-primitive distances leads to large or spurious cuboids occluding partsof the scene behind. We thus propose an occlusion-aware distance metriccorrectly handling opaque scenes. The proposed algorithm does not requirelabour-intensive labels, such as cuboid annotations, for training. Results onthe challenging NYU Depth v2 dataset demonstrate that the proposed algorithmsuccessfully abstracts cluttered real-world 3D scene layouts.

Cuboids Revisited: Learning Robust 3D Shape Fitting to Single RGB Images