Table 1 Summary of recent literature on census-independent population estimation. Note: ROI is region of interest, LOOCV is leave one out cross validation, MeAPEis median absolute percent error.
From: Census-independent population estimation using representation learning
ROI | Input resolution | Output resolution | Input data cost | Performance | ||
|---|---|---|---|---|---|---|
Validation | MeAPE | \(R^2\) | ||||
Kano and Kaduna, Nigeria15 | 0.5 m (Maxar) | 90 m | High (Maxar) | Survey (same region) | – | 0.98 |
Sri Lanka10 | 10 m (Object shape data) 12–30 m (Settlement Layer) 750 m (Night time Lights) | Village level | Free (Landsat) High (Maxar) | Train/test split | 28 | 0.58 |
Bo, Sierra Leone16 | 30 m (Landsat) | City district level | Free (public data) | LOOCV | 11 | – |
Nigeria11 | 0.5 m (Maxar), 100 m (WorldPop), various (OSM school density, household size) | 100 m | Free (OSM, WorldPop) High (Maxar) | Train/test split | – | 0.26 |
