Fig. 1: The Multi-task Observation using Satellite Imagery and Kitchen Sinks (MOSAIKS) approach transforms satellite imagery for each administrative polygon into a vector of image features. | Nature Communications

Fig. 1: The Multi-task Observation using Satellite Imagery and Kitchen Sinks (MOSAIKS) approach transforms satellite imagery for each administrative polygon into a vector of image features.

From: Global high-resolution estimates of the UN Human Development Index using satellite imagery and machine learning

Fig. 1: The Multi-task Observation using Satellite Imagery and Kitchen Sinks (MOSAIKS) approach transforms satellite imagery for each administrative polygon into a vector of image features.The alternative text for this image may have been generated using AI.

A The location of Oromia, an example province (ADM1 unit) within Ethiopia. B A composite of Planet imagery over Oromia in 2019. C A sample of 0.01∘ x 0.01∘ image tiles. D Three examples of MOSAIKS random convolutional features over Oromia; each pixel shows the feature value for a single 0.01∘ × 0.01∘ image (Xtile). E The corresponding aggregation of these MOSAIKS features to the provincial polygon (ADM1) level for model training (\({\overline{{{{\bf{X}}}}}}_{province}\)). F Aggregation of these same MOSAIKS features to the municipal polygon (ADM2) level for fine-resolution prediction of HDI (\({\overline{{{{\bf{X}}}}}}_{municipality}\)). See Figure S1 for an illustration of how MOSAIKS features are calculated and used to predict HDI. The municipal shapefile used here is from geoBoundaries65, published under a CC-BY 4.0 license.

Back to article page