Read our NeurIPS Paper
Our paper detailing our methodology to construct the dataset was published at the NeurIPS 2021 Datasets and Benchmarks Track.
Impacts of spatial apartheid
Analyzing the impacts of South African apartheid using computer vision techniques and satellite imagery. Read our NeurIPS paper and this MIT Tech Review article, and check out our dataset and visualizations.
What are the impacts of spatial apartheid?
Although apartheid was legally abolished in South Africa in 1991, its legacy remains today: the World Bank labels South Africa the world's most unequal country. To help researchers understand the current impact of spatial apartheid, we developed a dataset consisting of satellite imagery covering South Africa, accompanied by polygons labeled according to four classes of neighborhoods: wealthy areas, non wealthy areas, non residential neighborhoods and vacant land.
What we did
We developed a dataset and a classifier to study the evolution of spatial apartheid in South Africa. Our dataset was created using polygons demarcating land use, geographically labeled coordinates of buildings in South Africa, and high resolution satellite imagery covering the country from 2006-2017, and four classes of neighborhoods: wealthy areas, non wealthy areas, non residential neighborhoods and vacant land. This page is dedicated to resources on this work, including the dataset, visualizations, and other resources.
Download dataset
This dataset is freely available for academic and non-academic entities to use for non-commercial purposes such as academic research, teaching, scientific publications, or personal experimentation.
Request the dataset for download through this request form.
This work on Constructing a Visual Dataset to Study the Effects of Spatial Apartheid in South Africa was published at the NeurIPS 2021 Datasets and Benchmarks Track .
Click on the link to understand how we created the dataset.
Read the Datasheet
Read the datasheet for a detailed overview of the dataset, what it was created for, how we intend for it to be used, and what its limitations are.
Read the MIT Tech Review article on our work
Our spatial apartheid work was featured on MIT Tech Review: How satellite images and AI could help fight spatial apartheid in South Africa.
South Africa: 30 years after apartheid, what has changed?
Raesetje was interviewed by Al Jazeera in this in-depth look at the ways South African government's data practices obscure the true quality of life of township residents.
We have created a map where you can visualize the dataset below, with options to notify us if you believe that a neighborhood was incorrectly labeled.