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

Traditional attempts for loop closure detection typically use hand-craftedfeatures, relying on geometric and visual information only, whereas more modernapproaches tend to use semantic, appearance or geometric features extractedfrom deep convolutional neural networks (CNNs). While these approaches aresuccessful in many applications, they do not utilize all of the informationthat a monocular image provides, and many of them, particularly thedeep-learning based methods, require user-chosen thresholding to actually closeloops -- which may impact generality in practical applications. In this work,we address these issues by extracting all three modes of information from acustom deep CNN trained specifically for the task of place recognition. Ournetwork is built upon a combination of a semantic segmentator, VariationalAutoencoder (VAE) and triplet embedding network. The network is trained toconstruct a global feature space to describe both the visual appearance andsemantic layout of an image. Then local keypoints are extracted frommaximally-activated regions of low-level convolutional feature maps, andkeypoint descriptors are extracted from these feature maps in a novel way thatincorporates ideas from successful hand-crafted features. These keypoints arematched globally for loop closure candidates, and then used as a finalgeometric check to refute false positives. As a result, the proposed loopclosure detection system requires no touchy thresholding, and is highly robustto false positives -- achieving better precision-recall curves than thestate-of-the-art NetVLAD, and with real-time speeds.

CALC2.0: Combining Appearance, Semantic and Geometric Information for Robust and Efficient Visual Loop Closure