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

We propose simple yet effective improvements in point representations andlocal neighborhood graph construction within the general framework of graphneural networks (GNNs) for 3D point cloud processing. As a first contribution,we propose to augment the vertex representations with important local geometricinformation of the points, followed by nonlinear projection using a MLP. As asecond contribution, we propose to improve the graph construction for GNNs for3D point clouds. The existing methods work with a k-nn based approach forconstructing the local neighborhood graph. We argue that it might lead toreduction in coverage in case of dense sampling by sensors in some regions ofthe scene. The proposed methods aims to counter such problems and improvecoverage in such cases. As the traditional GNNs were designed to work withgeneral graphs, where vertices may have no geometric interpretations, we seeboth our proposals as augmenting the general graphs to incorporate thegeometric nature of 3D point clouds. While being simple, we demonstrate withmultiple challenging benchmarks, with relatively clean CAD models, as well aswith real world noisy scans, that the proposed method achieves state of the artresults on benchmarks for 3D classification (ModelNet40) , part segmentation(ShapeNet) and semantic segmentation (Stanford 3D Indoor Scenes Dataset). Wealso show that the proposed network achieves faster training convergence, i.e.~40% less epochs for classification. The project details are available athttps://siddharthsrivastava.github.io/publication/geomgcnn/

Exploiting Local Geometry for Feature and Graph Construction for Better 3D Point Cloud Processing with Graph Neural Networks