Open population datasets and open challenges

What this post added

This post details the creation and public release of high-resolution population datasets derived from satellite imagery using computer vision and convolutional neural networks. It explains the technical tradeoffs made in the model to achieve global scale and binary building detection, and the validation process with census data and household surveys. It also presents an analysis of the dataset using clustering algorithms to identify urbanized areas and determine population proximity to cities, which informs the design of communication technologies for expanding internet access.