IEEE OFCCT 2023: Keynote 3.4: The IEEE GRSS Earth Observation Database - Application Ready Data at Your Fingertips
IEEE Members: Free
Non-members: FreeLength: 01:11:47
Data-driven methods such as Deep Learning are the dominating approach in many scientific fields that require modeling a function mapping measurements to target variables. While by no means limited to it, this includes remote sensing and Earth observation with their applications to climate change, hazard management, monitoring and forecasting natural processes, and the sustainable development goals. However, learning-based systems can only be as good as the data they are trained on and thus require large datasets that provide both the actual measurements and values of the target variables. This presentation will discuss the current state and historical development of training datasets for machine learning in Earth observation and the role GRSS plays in their curation with a particular focus on how to find and access such datasets.