A Generalizable and Accessible Approach to Machine Learning with Global Satellite Imagery
This approach makes SIML more accessible for researchers in data-poor regions by reducing computational barriers, though it is incremental in improving efficiency.
The paper tackles the resource-intensive nature of combining satellite imagery with machine learning (SIML) by proposing a single encoding method that generalizes across diverse tasks like forest cover and house price prediction, achieving competitive accuracy with deep neural networks at orders of magnitude lower computational cost.
Combining satellite imagery with machine learning (SIML) has the potential to address global challenges by remotely estimating socioeconomic and environmental conditions in data-poor regions, yet the resource requirements of SIML limit its accessibility and use. We show that a single encoding of satellite imagery can generalize across diverse prediction tasks (e.g. forest cover, house price, road length). Our method achieves accuracy competitive with deep neural networks at orders of magnitude lower computational cost, scales globally, delivers label super-resolution predictions, and facilitates characterizations of uncertainty. Since image encodings are shared across tasks, they can be centrally computed and distributed to unlimited researchers, who need only fit a linear regression to their own ground truth data in order to achieve state-of-the-art SIML performance.