Analyzing and predicting urban land use forms in East Africa using OpenStreetMap data, satellite imagery, and Convolutional Networks

This multi-spectral satellite image data set is associated with our recent work on analyzing and predicting urban land use forms in East Africa using OpenStreetMap data, satellite imagery, and Convolutional Neural Networks.

The images were extracted using an automated Python script from Google Maps Static API, based on sample locations in four East African capital cities namely Kampala, Nairobi, Dar es Salaam, and Kigali.

Other data sets associated with this work, that is, ESRI shapefiles for administrative level 1 and OpenStreetMap data for the named cities may be downloaded directly from the respective URLs provided in the manuscript.

The data set is organized in 3 directories within the main directory namely train, validation, and test. Within each of these 3 directories are similarly named 10 directories containing images for each land use class. The directory names correspond to class names, e.g., directory name 'commercial' contains images representing commercial areas.