In this work, we introduce a new dataset of productive fields based on agriculture machine’s operation combined with Sentinel-2 images collected over a period of time, enabling to training highly accurate prediction models that automatically detect productive crop fields.