Title: Prediction accuracy of model

Description: The ResNET model shows a good precision at field level when Evapotranspiration and Soil Moisture are used as inputs. In this matrix, each cell shows the percentage of the correctly predicted labels over total of true labels for each irrigation type. For instance, considering the row of the not irrigated fields (the first one), 88.07% were identified correctly by the model, while the other 11.93% were mistakenly identified as drip irrigation. On the other hand, the first column on the left shows that some of the fields identified as non-irrigated were actually drip irrigated (4% of the total surface of drip irrigation). Overall, the model detected not irrigated fields with a 94.12% of precision.

Copyright: isardSAT & IRTA