Search for HOT projects, news, people and jobs.

Satellite Imagery for Social Good in Kenya and Nigeria

HOT Project


Satellite Imagery for Social Good is a single project with two distinct initiatives in Kenya and Nigeria. Teams in Kenya and Nigeria are using Map with AI to add and validate AI-generated building footprints into OpenStreetMap to generate data that decision-makers need to plan and implement social-economic programs. Implemented by the Humanitarian OpenStreetMap Team (HOT) in partnership with Microsoft.

The Satellite Imagery for Social Good project was developed and implemented in Kenya and Nigeria


The Humanitarian OpenStreetMap Team (HOT), Microsoft’s AI for Humanitarian Action, and Open Mapping Communities in Kenya and Nigeria have contributed and validated 1.4 million building footprints and 19,131 km of road in Nakuru, Turkana, and Kisumu. In Nigeria, 1.8 million buildings in the northern states of Bauchi and Gombe have been added to the map.

The AI-generated and human-verified building footprints have contributed to the development of city plans in Nakuru county, Kenya, and efficient planning for the provision of healthcare services in Nigeria.

In 2021, Microsoft provided a set of AI-generated building footprints to facilitate the importation of AI-generated, human-verified building footprints into OpenStreetMap (OSM). Thus, increasing the rate of remote mapping (tracing buildings, roads, and other features on satellite images using a simple web tool) and allowing mappers more time for validation and ground-truthing activities.


Project Use Cases


In Nakuru, HOT worked with the Nakuru City Board to generate building footprints, road data, and community-identified flood-prone areas in Nakuru City. Within a period of ten months, HOT trained 20 OSM members and six county and city planning officials on AI-enabled open mapping tools. The training included an introduction to OpenStreetMap’s mapping tools namely, RapID and JOSM in addition to data collection tools, ODK, OSMAnd, and StreetComplete, necessary to add extra attributes such as street names, schools, and other points of interest to the map.

WNA_ESA_SatelliteImageryforSocialGood_Nakuru City Board.jpeg

Satellite Imagery for Social Good video - Kenya

Urbanization and unsustainable environmental modification have significantly contributed to flooding in the lower parts of Nakuru City. The generated data sets from the project are currently being used by the council and city planners to assess flood risk and develop mitigation strategies using the sponge city model. Flood risk assessment in Nakuru City, part 1, and Flood risk assessment in Nakuru City part 2.



In Nigeria, the data generated (roads and building footprints) was combined with existing health datasets from the local government and key local health actors. Project partners are planning to use this data to improve the quality of micro planning and the overall reach of health campaigns by accurately estimating and locating underserved populations in Bauchi, Borno, and Gombe states. Over 35 stakeholders, including participants from eHealth Africa, fhi360, UNICEF, and WHO, and 19 members from local OSM communities participated in workshops, training, and remote mapping activities as part of the initiative. The training sessions included an introduction to open participatory mapping, working with mobile data collection apps, and an introduction to map creation and usage.


Bauchi State Primary Health Care Development Agency (BSPHCDA), is planning to use the data generated through the project to improve healthcare immunization programs and child protection initiatives in the northern state of Nigeria.

WNA_ESA_SatelliteImageryforSocialGood_bauchimap1.jpeg WNA_ESA_SatelliteImageryforSocialGood_bauchimap2.jpeg

Social Imagery for Social Good video - Nigeria

AI-assisted mapping contributed to greater accuracy in building geometry in several places, aided faster remote mapping, improved data quality, and reduced imagery offsets.

MapwithAI, JOSM provided community mappers with the opportunity to make the relevant amendments and validate the AI layer before adding the buildings to the OSM layer.

HOT developed a conflation software to implement a Manual Conflation Workflow across Tasking Manager’s implemented projects aiming to evaluate the quality of imported data. Additionally, a detailed analysis for the intersection of union (IoU) of the imported Microsoft AI open buildings dataset has been implemented with IoU score +90% overall.

Challenges experienced via the AI-generated data included instances of overlapping buildings identified as one unit resulting in an incorrectly mapped outline, false positives with other structures identified as buildings, and true negatives, for example, huts in rural areas were not identified as buildings and therefore were not mapped by AI.