Using computer vision, based on supervised machine learning, to support a global community of farmers
From crop diversity to weather patterns, farmers need to consider a high number of variables when deciding how to manage their fields. Accurately monitoring the growth of crops and the effects of various interventions can have a significant impact on yield and income. Gathering the relevant information can require costly specialist equipment and can be difficult to maintain over large territories. The ability to integrate multiple vast datasets in real time to generate predictions is also beyond the capabilities of individual farmers and their teams.
OneSoil is an app that has been developed to help farmers and their teams make better decisions. It uses machine‑learning and computer vision to assign field boundaries and crop types to farmland, as well as calculate a vegetation index as a proxy for crop health. The OneSoil algorithms are trained on the high-quality open data obtained through the European Commission’s satellite programme, Copernicus. On the platform, this data is combined with localised measurements from on-the-ground sensors tracking various environmental features – such as soil moisture, air humidity and temperature – to give a more comprehensive understanding of the state of every field. This up-to-date information collected by farmers through sensors provides the vital additional detail needed to accurately model the farming environment, predict outcomes and manage interventions.
OneSoil also supports effective team communication and collective learning. All employees, from machine operators to managers, can add information onto the platform, including correcting mislabelled crops and adding manual annotations and image attachments reflecting the current status of the fields.
The cross analysis of big data from weather sensors and satellite images, alongside active annotations from users, helps farmers predict their lands’ productivity, plan and monitor fieldwork and predict the emergence of possible plant diseases and pests. The maps on the platform are updated every two to five days, so farmers can manage their fields using near real-time data. OneSoil has mapped more than 35 million fields in Europe alone and is being actively used for decision‑making by the global agricultural community.
 The United Nations’ Food and Agriculture Organization, published the Handbook on climate information for farming communities (2019), which clearly states that "‘agrometeorological data must be collected at the local level in order to represent the farming environment realistically"’ –- also relevant for IBM weather forecast.