Thread by @geonanayi

1: To connect every school & to mitigate the educational resources, first we need to find every school. Using AI and satellite imagery we helped @UNICEFinnovate @Gigaconnect and @ProCoWorld put find 23,100 unmapped schools in eight countries across Asia, Africa, and South America

2: A few highlights throughout the projects 1) we’ve been using supertile to train #machinelearning that achieve higher model performance. Training model with #Kubernetes engines and #Kubeflow allowed us to run model training and experiments faster

3: 2) Our open-sourced tool, ML-Enabler allows us to run #AI model inference in eight countries within a week. This is an efficient way to scan through millions of tiles and look for given school features.

4: 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.

5: Using a human-in-the-loop #activelearning process is critical. @developmentseed's 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.

6: Read more of our blog post: developmentseed.org/blog/2021-03-1…;
Explore interactive maps: devseed.com/project-connec…;
Example of unmapped schools: paper.dropbox.com/doc/Project-Co…