Governments and humanitarian communities rely on open and accessible maps to provide relief and development assistance to every community in need. Today the increasing availability of high-resolution satellite imagery along with breakthroughs in cloud computing and computer vision has allowed remote households to be included in recognizable maps and also swiftly update datasets. These emerging technologies offer significant potential for humanitarian purposes – especially during a crisis.
High resolution satellite imageries allow accurate delineation of buildings. However still accessing timely, accurate data at a household level remains a challenge in many low-and-middle-income countries. Such data gaps can remain unaddressed, and humanitarian decisions are made with major information asymmetries across communities.
The DevGlobal Team is preparing to launch a program named RAMP (Replicable AI for Microplanning) this September. The project, to be launched in partnership with the World Health Organization (WHO), is an open-source, deep learning AI model program, and aims to democratize AI for use in humanitarian applications by lowering the bar for entry into automated building extraction using satellite imagery. Designed for flexibility and accessibility, the model has many uses, including emergency and climate change response.
Headquartered in Edmonds, Washington State, US, DevGlobal supports the world’s leading non-profits, companies, philanthropies, multilateral organizations, and government agencies on their innovation and sustainability initiatives.
“The project has focused initially on priority global health use cases and sees broad applicability across many domains – both commercial and societal benefit applications. This includes providing valuable baseline datasets for many development domains from energy planning to financial services access to equitable education,” says Rhiannan Price, Managing Director of DevGlobal and Project Manager for RAMP.
Open, accessible, impactful AI
The RAMP project’s primary aspiration is to democratize and empower humanitarian agencies to shape their strategies better by turning over control of the data value chain to them.
The Replicable AI can be used for digitizing buildings in low-and-middle-income countries using satellite imagery as well as enable in-country users to build their own deep learning models for their regions of interest, Price adds.
The project is to create digital micro plans, which are detailed maps depicting primary health center catchment areas for service delivery planning.
RAMP’s building footprints
The model is aimed at a wide range of user applications in times to come. The most important outreach is planned to cover effective provision of goods and services.
While RAMP’s microplanning use case primarily includes the tracking of the delivery of vaccines for the ongoing pandemic response, it also extends to other health programs as well as planning of rural electrification or roads or cash assistance and much more. Building-level datasets are an important input for planning and delivery of services to every household.
Risk exposure assessment is another area where RAMP model can be applied. Knowing where people live and work is critical to assessing current and future risk. Having data on human footprint is necessary to truly understand their vulnerabilities and risks and possible scenarios.
Validation of existing data including the ability to validate locations and improve existing data to make it more interoperable and analyzable with other datasets.
Population density is one of the most important statistics for development efforts across many sectors; however, many countries still lack complete, up-to-date census data. Building counts and locations allow modelers and statisticians to accurately estimate and distribute populations.
Such a model can empower mapping till the last building is incorporated and this would help increasing the accuracy in population density assessment.
Sampling for household surveys and mobile data collection campaigns, especially in rural areas where obtaining information on household location is difficult may be more organized with this model application. Individual building footprints are valuable, and given how rapid growth of communities are happening, being able to update those buildings data layers with an open-source model is perhaps a sustainable solution.
Support for any future pandemic
Deploying limited resources and services effectively and quickly is paramount to saving lives during any crisis.
“Throughout the pandemic, a lack of detailed data on where communities are and how to reach them delayed decision-making and called into question the ability to effectively respond and balance the needs of many countries,” Price points out.
“RAMP’s accessible machine learning model gives humanitarians the ability to produce mission-critical data on demand. By democratizing access to these methods, RAMP is ensuring that in-country experts have the data they need in a crisis. Knowing how much vaccine is needed or where it needs to be transported are basic questions that tools like RAMP can help answer,” she explains.
These foundational use cases would create a strong starting point enabling further use cases to follow. For instance, users can further innovate using RAMP’s codebase and data to build out even more robust analytics tailored to local needs, says Price. Such local usage can range from whether it’s for understanding where to send mobile clinics or community risk profiles or mapping buildings as a fundamental first step to precision health.
The RAMP model will enable a wide range of applications in times to come and help create more effective maps. Democratizing artificial intelligence definitely enables maps to facilitate wider community empowerment.
