High resolution images were downloaded from Bing or Google at a resolution of 0.5 metres per pixel. The images are georeferenced and stored in tiles, each tile covering a quarter-degree square (about 27 x 27 kilometres) and named after the quarter degree square block reference code. The images were systematically scoured to look for households using criteria listed below. Individual households were mapped using ArcView or QGis when they were isolated or clustered in small villages, while larger villages were mapped as polygons, their borders following the margins of areas in which houses were clustered. Field data collected by Panthera for the survey of large carnivores were used to delete houses and villages that Panthera found to be abandoned. The Panthera survey was conducted in 2016 and 2017. Individual households and polygon borders were also mapped in a sample of villages and urban areas to generate a regression equation to predict the number of household in areas mapped only as polygon areas. The resulting equation was: y = 0.0015x (where x is the polygon area in square metres). The y intercept was fixed at 0, and the r2 for 181 areas mapped as both point households and polygons was 0.876.