Image Quality Labels for Land Cover Datasets

To improve the understanding of data uncertainty in land cover mapping image quality labels are created for subsets of existing datasets by manual annotation. The first subset contains labels for the ISPRS Potsdam dataset (https://www.isprs.org/resources/datasets/benchmarks/UrbanSemLab/2d-sem-label-potsdam.aspx) where effects related to the orthorectification process are marked as low quality, i.e. artifacts, ghost effects, and blur. The second subset contains labels for the VALID dataset (https://sites.google.com/view/valid-dataset/) where dark shadows possibly impact the class assignment. Classes: 0 - Background 1 - Image degradation