Uncertainty Quantification and Propagation for Land Cover Classification in Remote Sensing

Kurzfassung

The rapid rise of Deep Learning (DL) has thoroughly transformed the field of Remote Sensing (RS). However, as these models have become increasingly complex, the fundamental need for Uncertainty Quantification (UQ) has emerged to provide a deeper understanding and more reliable interpretation of the models’ predictions. This dissertation addresses the critical need for UQ in DL models for RS applications, particularly in Land Use / Land Cover classification.