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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

This paper describes Georeference Contrastive Learning of visualRepresentation (GeoCLR) for efficient training of deep-learning ConvolutionalNeural Networks (CNNs). The method leverages georeference information bygenerating a similar image pair using images taken of nearby locations, andcontrasting these with an image pair that is far apart. The underlyingassumption is that images gathered within a close distance are more likely tohave similar visual appearance, where this can be reasonably satisfied inseafloor robotic imaging applications where image footprints are limited toedge lengths of a few metres and are taken so that they overlap along avehicle's trajectory, whereas seafloor substrates and habitats have patch sizesthat are far larger. A key advantage of this method is that it isself-supervised and does not require any human input for CNN training. Themethod is computationally efficient, where results can be generated betweendives during multi-day AUV missions using computational resources that would beaccessible during most oceanic field trials. We apply GeoCLR to habitatclassification on a dataset that consists of ~86k images gathered using anAutonomous Underwater Vehicle (AUV). We demonstrate how the latentrepresentations generated by GeoCLR can be used to efficiently guide humanannotation efforts, where the semi-supervised framework improves classificationaccuracy by an average of 10.2% compared to the state-of-the-art SimCLR usingthe same CNN and equivalent number of human annotations for training.

GeoCLR: Georeference Contrastive Learning for Efficient Seafloor Image Interpretation