A Weakly Supervised Approach for Estimating Spatial Density Functions from High-Resolution Satellite Imagery

TL;DR

This paper introduces a neural network component that enables training pixel-level density estimators from coarse aggregate data, improving high-resolution density estimation from satellite imagery without domain-specific assumptions.

Contribution

A novel regional aggregation layer allowing weakly supervised training of density estimators using only aggregate data, applicable to satellite imagery analysis.

Findings

Outperforms baseline density estimation methods.

Effective in estimating population and housing densities.

Enables building classification based on density estimates.

Abstract

We propose a neural network component, the regional aggregation layer, that makes it possible to train a pixel-level density estimator using only coarse-grained density aggregates, which reflect the number of objects in an image region. Our approach is simple to use and does not require domain-specific assumptions about the nature of the density function. We evaluate our approach on several synthetic datasets. In addition, we use this approach to learn to estimate high-resolution population and housing density from satellite imagery. In all cases, we find that our approach results in better density estimates than a commonly used baseline. We also show how our housing density estimator can be used to classify buildings as residential or non-residential.