Accurate and comprehensive measurements of a range of sustainable development\noutcomes are fundamental inputs into both research and policy. We synthesize\nthe growing literature that uses satellite imagery to understand these\noutcomes, with a focus on approaches that combine imagery with machine\nlearning. We quantify the paucity of ground data on key human-related outcomes\nand the growing abundance and resolution (spatial, temporal, and spectral) of\nsatellite imagery. We then review recent machine learning approaches to\nmodel-building in the context of scarce and noisy training data, highlighting\nhow this noise often leads to incorrect assessment of models' predictive\nperformance. We quantify recent model performance across multiple sustainable\ndevelopment domains, discuss research and policy applications, explore\nconstraints to future progress, and highlight key research directions for the\nfield.\n