Towards Sustainable Census Independent Population Estimation in Mozambique

TL;DR

This paper explores sustainable, census-independent population estimation in Mozambique using remote sensing, microcensus data, transfer learning, and publicly available datasets to improve accuracy and feasibility.

Contribution

It introduces a method combining transfer learning and publicly available data for sustainable population estimation in Mozambique.

Findings

Population predictions improve with footprint area estimation.

Using publicly available datasets is feasible for sustainable estimation.

Transfer learning enhances building footprint predictions.

Abstract

Reliable and frequent population estimation is key for making policies around vaccination and planning infrastructure delivery. Since censuses lack the spatio-temporal resolution required for these tasks, census-independent approaches, using remote sensing and microcensus data, have become popular. We estimate intercensal population count in two pilot districts in Mozambique. To encourage sustainability, we assess the feasibility of using publicly available datasets to estimate population. We also explore transfer learning with existing annotated datasets for predicting building footprints, and training with additional `dot' annotations from regions of interest to enhance these estimations. We observe that population predictions improve when using footprint area estimated with this approach versus only publicly available features.