BlueGreen Labs is proud to present some of our recent work in collaboration with ACRE Africa, IFPRI and supported by the LACUNA fund on publishing a formal machine learning ready data set on Radiant Earth within the context of the Eyes on The Ground project.
The data set, and Spatio-Temporal Asset Catalog (STAC), contains almost 30K images of crop fields made by farmers through a dedicated cellphone application. These images give a proxy for the state of the crop at a given point in time. Image data are matched with ancillary data such as Sentinel-2 vegetation indices, ERA5 climate data and various proxies for daily rainfall (CHIRPS, ARC2). In addition, labels are provided of crop disturbances as classified by agronomists.
Although exact field locations are hidden due to privacy reasons, and only village level centroids are reported, this data should allow for the estimation of (disturbance) impacts based upon the data the provided climatology and data labels.
You can explore the data on the Radiant Earth Machine Learning Hub: