The Ablation Study: Irrigation Detection In Satellite Images using Deep Learning

The BigEarthNet dataset consists of 590,326 Sentinel-2 satellite images, collected for the purpose of remote sensing. To our knowledge, while many researchers have used the dataset, none have done an ablation study to systematically establish the interactions between feature selection and predictive model performance. Our group seeks to fill that gap in knowledge in an effort to establish a foundation for future researchers who intend to use the dataset for studies related to agriculture and irrigation.