3 ms·
Fighting Poverty with Satellite Images and Machine Learning Wizardry
- Aelinsaar 10y agoFinally, a way to assess poverty from space, ironically increasing the distance between the observers and the observed by several literal orders of magnitude.
- jostmey 10y agoI guess people just don't want to believe that poverty can exist in their neighborhood. They have to believe it is somewhere far away on the other side of the planet.
- Houshalter 10y agoFirst world poverty is bad, but it's nothing compared to the worst of third world poverty. By any measure the poor of the first world are significantly better off than the third world. Income, life expectancy, access to luxuries like clean and running water to artificial lighting, etc.
- eli_gottlieb 10y agoAnd Third World poverty is kept at bay in the First World only by ever-vigilant effort amongst government and civil society.
- Houshalter 10y agoTo some extent yes. I think even without social programs they would still be better off, as there are no social programs in the third world. Wages are just vastly higher in the first world, even for the poorest. Third worlders live on a dollar a day or less. Social programs help a lot of course.
- cuchoi 10y agoThe abstract of the paper was more informative than the article: "Reliable data on economic livelihoods remain scarce in the developing world, hampering efforts to study these outcomes and to design policies that improve them. Here we demonstrate an accurate, inexpensive, and scalable method for estimating consumption expenditure and asset wealth from high-resolution satellite imagery. Using survey and satellite data from five African countries—Nigeria, Tanzania, Uganda, Malawi, and Rwanda—we show how a convolutional neural network can be trained to identify image features that can explain up to 75% of the variation in local-level economic outcomes. Our method, which requires only publicly available data, could transform efforts to track and target poverty in developing countries. It also demonstrates how powerful machine learning techniques can be applied in a setting with limited training data, suggesting broad potential application across many scientific domains." http://science.sciencemag.org/content/353/6301/790 http://science.sciencemag.org/content/353/6301/790
- gcr 10y agoThis seems quite relevant to the work that Orbital Insight is doing.