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

3D object detection and dense depth estimation are one of the most vitaltasks in autonomous driving. Multiple sensor modalities can jointly attributetowards better robot perception, and to that end, we introduce a method forjointly training 3D object detection and monocular dense depth reconstructionneural networks. It takes as inputs, a LiDAR point-cloud, and a single RGBimage during inference and produces object pose predictions as well as adensely reconstructed depth map. LiDAR point-cloud is converted into a set ofvoxels, and its features are extracted using 3D convolution layers, from whichwe regress object pose parameters. Corresponding RGB image features areextracted using another 2D convolutional neural network. We further use thesecombined features to predict a dense depth map. While our object detection istrained in a supervised manner, the depth prediction network is trained withboth self-supervised and supervised loss functions. We also introduce a lossfunction, edge-preserving smooth loss, and show that this results in betterdepth estimation compared to the edge-aware smooth loss function, frequentlyused in depth prediction works.

VR3Dense: Voxel Representation Learning for 3D Object Detection and Monocular Dense Depth Reconstruction