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 Tools annual

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 monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Geographic location is essential for modeling tasks in fields ranging fromecology to epidemiology to the Earth system sciences. However, extractingrelevant and meaningful characteristics of a location can be challenging, oftenentailing expensive data fusion or data distillation from global imagerydatasets. To address this challenge, we introduce Satellite ContrastiveLocation-Image Pretraining (SatCLIP), a global, general-purpose geographiclocation encoder that learns an implicit representation of locations fromopenly available satellite imagery. Trained location encoders provide vectorembeddings summarizing the characteristics of any given location for convenientusage in diverse downstream tasks. We show that SatCLIP embeddings, pretrainedon globally sampled multi-spectral Sentinel-2 satellite data, can be used invarious predictive tasks that depend on location information but notnecessarily satellite imagery, including temperature prediction, animalrecognition in imagery, and population density estimation. Across tasks,SatCLIP embeddings consistently outperform embeddings from existing pretrainedlocation encoders, ranging from models trained on natural images to modelstrained on semantic context. SatCLIP embeddings also help to improve geographicgeneralization. This demonstrates the potential of general-purpose locationencoders and opens the door to learning meaningful representations of ourplanet from the vast, varied, and largely untapped modalities of geospatialdata.

SatCLIP: Global, General-Purpose Location Embeddings with Satellite Imagery