Ramp Use Cases

How will Ramp be used to help address real-world problems?

Our team performed outreach and collaborated with geospatial partners to develop use case vignettes demonstrating Ramp’s scalability and applications for a variety of geospatial project outcomes. The team will continue to build and document use cases that highlight benefits for end users of AI-generated building footprints, as well as Return on Investment for governments, humanitarian partners and others.

Center for Remote Sensing and Geographic Information Services (CERSGIS), University of Ghana

Improving flood risk, solid waste management, and access to basic infrastructure

CERSGIS is a non-profit entity with the mission to support development planning with geo-information technologies, generate geo-information products for decision making, build capacity, and make geospatial data widely available for research and development.

CERSGIS provided map products for the Environmental and Social Impact Assessment (ESIA) of the Greater Accra Resilient and Integrated Development project. The goal of the project is to improve flood risk and solid waste management in the Greater Accra Metropolitan Area and improve access to basic infrastructure and services in targeted communities within the major river basin covering the area. The availability of high-resolution building footprint data facilitated the identification of hot spots and vulnerable populations subjected to perennial flooding.

This project is focused on progressive reduction of flooding and related risk of losing lives, assets, and economic opportunities in the vulnerable communities. Map products helped in the detailed design of drains, access roads, alleyways, streetlights, and solid waste management facilities in flood-prone communities. Also, the map products are helping in the implementation of community-based solid waste management interventions in targeted crowded communities and outreach programs to sensitize and improve public behavior on solid waste management and litter management.

The map products used for the ESIA provide a “snapshot model” of the project area. Hopefully, the Ramp model will provide near real-time data for the development of a monitoring system for the impact assessment studies.

Humanitarian OpenStreet Map Team (HOT)

Training models based on feedback from local OSM mappers

The Humanitarian OpenStreetMap Team (HOT) envisions a world where community needs are addressed through mapping, anyone can access and contribute to the map, and data is available and used for impact. HOT’s Technology & Innovation unit (hot_tech team) exists to amplify connections between humanitarian needs and open map data by pursuing just and fair tech. By just, we mean designing and building technologies with intention, investing in people and practices to ensure they do less harm and more good. To ensure fair technologies, we put humanitarian needs at the center of their free and open source software (FOSS) technologies. Some important principles in this quest are ensuring we receive feedback from end-users, keep the AI models open, and remain aware of model bias.

We see that mappers can, on average, map between 1000-1500 buildings per working day without AI assistance. During an AI-assisted mapping pilot (2019-2020) supported by Microsoft, 18 million building footprints were extracted from satellite imagery for all of Tanzania and Uganda. HOT discovered during this pilot that this average nearly doubled to 2500-3000 buildings being added to OpenStreetMap (OSM) per day with the assistance of high-quality AI models. This created a proof of concept for developing inclusive, just and open source AI models that are integrated in a workflow to assist OSM mappers and map communities more efficiently and precisely. We call it fAIr

fAIr is a service that uses AI models to detect map features based on satellite imagery and add them into OpenStreetMap. Unlike other AI data producers, fAIr is a free and open-source AI service that offers local communities accurate  feedback through the efforts of OSM community mappers. This results in progressive intelligence of computer vision models. Whenever an OSM mapper uses the AI models for assisted mapping and completes corrections (which are the weaknesses of the models), fAIr can take those corrections as feedback to enhance the model’s accuracy. 

In the example below, the left image is the model prediction which identified empty land as a building and missed 2 building boundaries. The OSM user then fixes those issues and pushes them to OSM. Next, fAIr can take those corrected features and teach the AI model about its misses.

fAIr is at an exciting stage in its lifecycle where the models are being trained based on feedback from local OSM mappers. This eliminates model biases as we ensure the models are relevant to the communities where the maps exist to improve the conditions of the people living there. Our working goal is to provide OSM mappers access to AI-assisted mapping across mobile and in-browser editors using community-created AI models.

The output of Ramp would be an open-source AI model that we can integrate into fAIr. There are three components in fAIr:

  1. The training datasets would provide input for model training.
  2. Open source models such as the RAMP model can be re-trained seeking a progressive increase in intelligence (accuracy), without an AI engineer’s intervention.
  3. Prediction will run live and be presented to end users for validation and confirmation.

Here you can see the three components of fAIr.

Clinton Health Access Initiative (CHAI)

Determining sub-villages

CHAI is a global health organization committed to saving lives and reducing the burden of disease in low-and middle-income countries. CHAI works with partners to strengthen the capabilities of governments and the private sector to create and sustain high-quality health systems that can succeed without third party assistance. In addition to supporting disease prevention, women and children’s health, universal coverage and other cross-cutting health areas, CHAI is working to ensure that technology is used effectively to catalyze government health goals by collaborating with governments, developers, donors, and end users (health care providers, health systems managers, and patients) to design, develop, scale, and institutionalize fit-for-purpose digital products.

Digital health facility catchment maps, paired with physical accessibility analyses, provide key contextual information on the geographic access to immunization services, identifying areas which are hardest to reach during microplanning. Though high-quality health facility, stamp village (villages with a village chief), transport network and terrain data are available in Laos to form as a basis of such maps, there is incomplete information on the number and location of smaller sub-villages, which may be home to a considerable number of households. These communities may therefore be missed in rounds of microplanning, as these are limited to the stamp village level.

Having access to building footprints can serve as an easy and accurate way to identify clusters of households – and therefore the location of sub-villages – in health facility catchment maps and ensure these smaller communities are included when planning immunization campaigns. One major challenge with this approach, however, is having stakeholders at various levels (from central down to health center) agree on the definition of a sub-village (i.e. what actually constitutes a sub-village? 10 households? 20 households? What about informal settlements such as camps?), both within and across programs, the latter being important when it comes to integrated health interventions. There may therefore need to be some work done on the governance side, for there to be at least within the program a single definition of a sub-village.

Ramp can be trained on Laos satellite imagery to generate building footprints that are subsequently analyzed to determine clusters of households that can be considered sub-villages.


Microplanning, malaria bednet distribution, disaster response and infrastructure planning efforts

Akros is a cutting-edge organization that establishes data-driven systems to improve the health and well-being of disadvantaged communities. We pride ourselves in our ground-level knowledge of the service delivery systems where we work, and our ability to provide novel, lasting solutions implemented in developing regions.

Reveal is an open-source platform that uses spatial intelligence to drive delivery of live-saving interventions using program-specific geospatial database to improve map use and data for microplanning. When combined with Ramp, these two Digital Public Goods have the ability to accelerate the pace at which new regions are mapped and field activities are planned and executed.

Before the integration of Ramp-produced data into Reveal can be deemed production-ready, there is an immediate need to pressure test the data workflow and assess the accuracy of the footprints that inform the given health campaign, which will constitute Phase 1 of our collaborative efforts. Reveal has already been used in numerous malaria and neglected tropical disease (NTD) programs across West, Southern and Eastern Africa to achieve immediate impact, and as such the tool has generated data that can be used to validate Ramp-derived building footprints. The previously validated Reveal data will be compared to Ramp-produced building data, ensuring that our future plans of deploying Ramp in new geographies and visualizing in Reveal have the necessary accuracy and usability to be leveraged in global health campaigns.

The Ramp model can be re-trained and deployed on new geographies to produce building data significantly faster than manual digitization that will be more up-to-date than common building data sources. On-demand update of building footprint datasets can have immediate positive impact for use cases including microplanning, malaria bednet distribution, disaster response planning and infrastructure planning efforts.