Currently we are in East Africa working on one of our larger jobs — a Gates Foundation project called AfSIS, which stands for Africa Soil Information Service. The project aims to build accurate GIS maps of Africa’s soil resources through precision agriculture techniques, and in so doing, we hope to ultimately increase the prosperity of rural communities and nations. Soil properties such as Nitrogen, Carbon, and pH are estimated across the continent using our spatial-temporal machine learning models, which are fed with datastreams funneled in from wet chemistry and spectroscopy labs, field measurements, and satellites. While our data and users often come from impoverished areas, processing this volume of data necessitates using very modern techniques, such as cloud computing on clusters with thousands of instances, and converting all algorithms from high-level languages (e.g. R/KNIME/MATLAB) into efficient C++.
As shown in the video above, recently we started experimenting with UAVs, with the hope of integrating high-resolution imagery (e.g. ~5cm) from customized field surveys into our system, and also running computer vision algorithms on the feeds. Below are some videos of test flights executed in the field, taken from Google Glass.