Using Machine Learning to Build a More Food Secure Future

Food insecurity remains an urgent development challenge: 750 million people worldwide are severely food insecure, and climate change is causing food production to be less predictable. Machine learning and earth observation are helping to guide the interventions of NGOs like the World Food Programme. However, in many countries in the Global South, a scarcity of labeled data prevents the successful application of these methods. Frontier Development Lab has been working with the World Food Programme to develop self-supervised embeddings in data-scarce countries, significantly enhancing the available information in these countries and reducing the time and cost overheads of field surveys. We have built a scalable data pipeline to obtain years' worth of Sentinel-2 imagery over hundreds of thousands of square kilometers in mere minutes. Our self-supervised machine learning model produces embeddings which we are using to map the foodscape and even classify crop types.

WATCH LIVE ON YOUTUBE: https://youtu.be/j3_Foi8Qv8Y
WATCH LIVE ON FACEBOOK: https://fb.me/e/RYe9bpwd