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

With the increasing reliance of self-driving and similar robotic systems onrobust 3D vision, the processing of LiDAR scans with deep convolutional neuralnetworks has become a trend in academia and industry alike. Prior attempts onthe challenging Semantic Scene Completion task - which entails the inference ofdense 3D structure and associated semantic labels from "sparse" representations- have been, to a degree, successful in small indoor scenes when provided withdense point clouds or dense depth maps often fused with semantic segmentationmaps from RGB images. However, the performance of these systems dropdrastically when applied to large outdoor scenes characterized by dynamic andexponentially sparser conditions. Likewise, processing of the entire sparsevolume becomes infeasible due to memory limitations and workarounds introducecomputational inefficiency as practitioners are forced to divide the overallvolume into multiple equal segments and infer on each individually, renderingreal-time performance impossible. In this work, we formulate a method thatsubsumes the sparsity of large-scale environments and present S3CNet, a sparseconvolution based neural network that predicts the semantically completed scenefrom a single, unified LiDAR point cloud. We show that our proposed methodoutperforms all counterparts on the 3D task, achieving state-of-the art resultson the SemanticKITTI benchmark. Furthermore, we propose a 2D variant of S3CNetwith a multi-view fusion strategy to complement our 3D network, providingrobustness to occlusions and extreme sparsity in distant regions. We conductexperiments for the 2D semantic scene completion task and compare the resultsof our sparse 2D network against several leading LiDAR segmentation modelsadapted for bird's eye view segmentation on two open-source datasets.

S3CNet: A Sparse Semantic Scene Completion Network for LiDAR Point Clouds