The Lacuna Fund: Mobilizing Satellite Imagery Datasets for Better Machine Learning in Agriculture
Lacuna Fund’s first request for proposals (RFP) to support labeled datasets for agriculture in sub-Saharan Africa is open and closes 3 September. See the full announcement below or read more at http://www.lacunafund.org.
Machine learning methods, in conjunction with satellite remote sensing imagery, show enormous potential to benefit agriculture, allowing policymakers to access better information to drive policy making and farmers to receive tailored advice and predictions that can improve livelihoods. However, the labeled satellite imagery data that fuels these machine learning tools is often non-existent, biased, or private.
The Rockefeller Foundation, Google.org, Canada’s International Development Research Centre, and the German development agency GIZ on behalf of the Federal Ministry for Economic Cooperation and Development (BMZ) have partnered with technical experts, affected stakeholders, and end users to create the Lacuna Fund, an initiative that will mobilize resources to fund open source, labeled datasets to solve urgent problems in low- and middle-income contexts globally. Meridian Institute serves as the secretariat and fiscal agent for the Fund.
Machine Learning and Earth Observation for Agriculture
Machine learning (ML)-enabled tools in agriculture have great promise to increase production and resilience and contribute to broader sustainable development goals. As Hamed Alemohammad, Chief Data Scientist at Radiant Earth Foundation and a member of Lacuna Fund’s Technical Advisory Panel for agriculture, explains, “[Earth observation] satellites capture data at a global scale, and ML techniques can use these data to map croplands at local, regional, and continental levels, which provide input for farmers and policymakers alike. In particular, the ability to estimate crop yield or detect pest/disease damage during the growing season will be game-changing in addressing food insecurity problems.”
Machine learning techniques rely on labeled data to build these tools. For example, satellite imagery linked to ground-truth field boundary or crop type information might allow a data scientist to create accurate crop type maps for a far greater area using ML and remote sensing. However, in many cases, the data required to build ML applications for real-world problems doesn’t exist. And where it does exist, it’s often outdated, private, or it’s not representative of underserved populations. Machine learning tools then “learn” these biases, which can lead to inaccurate or harmful outcomes. As Hamed puts it, if models are built using data from other regions because better data isn’t available, “[t]he results are either biased or plain wrong. Therefore, to improve the accuracy of ML models from satellite imagery worldwide, regionally curated training data is essential.”
Lacuna Fund is the world’s first collaborative effort to directly address this problem. A lacuna is a gap, a blank space, or a missing part. Data scientists, funders, and social entrepreneurs around the world have recognized missing information from many regions as a lacuna in labeled datasets and are committed to closing these gaps in order to reduce bias and improve accuracy.
The Fund currently has a two-year horizon, and the first two RFPs will focus on agriculture and language portfolios in Africa.
Our hope is that the Fund, and the open source datasets it produces, will elevate the work of data labeling and bring more attention to the essential role it can play in international development efforts and increasing access to the benefits of artificial intelligence for all. In doing so, we hope to rally prospective funders and catalyze future funding.
Join us and get notified of future opportunities at http://www.lacunafund.org.
Announcement of Request for Proposals
Lacuna Fund, a collaborative effort to mobilize datasets for machine learning that solve urgent problems in low- and middle-income contexts globally, has issued a request for proposals (RFP) for labeled datasets in crop and animal agriculture in sub-Saharan Africa.
The full RFP and more details on eligibility, selection criteria, as well as information about the Fund and upcoming calls, can be found at http://www.lacunafund.org/apply.
- We are seeking applications to create, expand, or maintain labeled datasets from organizations and partnerships with technical expertise in agricultural data collection and labeling. Proposals should also demonstrate a strong understanding of the machine learning landscape and the needs of end users.
- Applicants must be headquartered in Africa or have a substantial partnership with an entity headquartered in Africa. Lacuna Fund encourages collaboration between organizations to assemble a competitive proposal.
- Proposals will be selected by a Technical Advisory Panel based on the Fund’s principles: transformative potential, quality, accessibility, equity, ethics, and a participatory approach.
- Lacuna Fund values a collaborative and locally driven approach to data creation, expansion, and maintenance. We recognize that the continued usefulness and maintenance of open data will thrive in a community that is collectively invested in that data.
- While the Request for Proposals outlines some data needs identified by our Technical Advisory Panel, proposals are not restricted to these areas, and we welcome other ideas within the domain area that have a clearly articulated benefit.
Read more about Lacuna Fund’s open RFP in agricultural labeled data and apply on our website here. You can also sign up to receive future notifications of available funding.
Lacuna Fund is a funder collaborative between The Rockefeller Foundation, Google.org, Canada’s International Development Research Centre, and the German development agency GIZ on behalf of the Federal Ministry for Economic Cooperation and Development (BMZ). The Fund is governed by a multi-stakeholder steering committee composed of technical experts, thought leaders, local beneficiaries, and end users. Collectively, the Fund’s stakeholders are committed to creating and mobilizing labeled datasets that both solve urgent local problems and lead to a step change in machine learning’s potential worldwide.