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

Cross-view geo-localization aims to estimate the location of a query groundimage by matching it to a reference geo-tagged aerial images database. As anextremely challenging task, its difficulties root in the drastic view changesand different capturing time between two views. Despite these difficulties,recent works achieve outstanding progress on cross-view geo-localizationbenchmarks. However, existing methods still suffer from poor performance on thecross-area benchmarks, in which the training and testing data are captured fromtwo different regions. We attribute this deficiency to the lack of ability toextract the spatial configuration of visual feature layouts and models'overfitting on low-level details from the training set. In this paper, wepropose GeoDTR which explicitly disentangles geometric information from rawfeatures and learns the spatial correlations among visual features from aerialand ground pairs with a novel geometric layout extractor module. This modulegenerates a set of geometric layout descriptors, modulating the raw featuresand producing high-quality latent representations. In addition, we elaborate ontwo categories of data augmentations, (i) Layout simulation, which varies thespatial configuration while keeping the low-level details intact. (ii) Semanticaugmentation, which alters the low-level details and encourages the model tocapture spatial configurations. These augmentations help to improve theperformance of the cross-view geo-localization models, especially on thecross-area benchmarks. Moreover, we propose a counterfactual-based learningprocess to benefit the geometric layout extractor in exploring spatialinformation. Extensive experiments show that GeoDTR not only achievesstate-of-the-art results but also significantly boosts the performance onsame-area and cross-area benchmarks.

Cross-view Geo-localization via Learning Disentangled Geometric Layout Correspondence