UNICEF and Development Seed are working to leverage machine learning, high-resolution imagery, and inexpensive cloud computing to create a comprehensive map of school at a global scale. Accurate data about school locations is critical to provide quality education and promote lifelong learning, UN sustainable development goal 4 (SDG4), to ensure equal access to opportunity (SDG10) and eventually, to reduce poverty (SDG1). However, in many countries, educational facilities’ records are often inaccurate or incomplete. Understanding the location of schools can help governments and international organizations gain critical insights into the needs of vulnerable populations, and better prepare and respond to exogenous shocks such as disease outbreaks or natural disasters. Unfortunately, some national governments still don’t know where all the schools in their country are or have out-of-date school maps.

Despite their varied structure, many schools have identifiable overhead signatures that make them possible to detect in high-resolution imagery with deep learning techniques. Approximately 18,000 previously unmapped schools across 5 African countries, Kenya, Rwanda, Sierra Leone, Ghana, and Niger, were found in satellite imagery with a deep learning classification model. These 18,000 schools were validated by expert human mapping analysts. In addition to finding previously unmapped schools, the models were able to identify already mapped schools with accuracy between 77 - 95% depending on the country. To facilitate running model inference across over 71 million zoom 18 tiles of imagery development seed relied on our open-source tool ML-Enabler.
ML Enabler generates and visualizes predictions from models that are compatible with Tensorflow’s TF Serving. ML-Enabler makes managing the infrastructure for running inference at scale and visualizing predictions straight-forward from a UI. ML Enabler will spin up the required AWS resources and run inference to generate predictions. ML Enabler helps harness the power of expert human mappers because model predictions can be validated within the UI and validated predictions can be used to generate new training data and re-train the initial model.
However, a notable limitation of our approach is it relies on human validators for both the training data creation and school validation. As a result, we may introduce a bias for schools that follow common patterns that are recognizable from space. Using a human-in-the-loop and active learning process is critical.
At Development Seed we are engaged in active research and development of human-in-the-loop active learning methods that allow non-expert human mappers and AI to work more efficiently together and improve model prediction.