The microwave data benchmark for classifi-
cation only reaches an accuracy of 0.8 which means that the two data sets do not
fully agree on melt areas. For our project, this means that we have to trust that the
optical data set is superior to the microwave data set in its accuracy. If this assump-
tion holds true, we can use the optical data as labels and we can improve melt map
precision. But we need to investigate this matter further by checking the overlap of
the two data sets closer, especially around the melt boundary.
The microwave data benchmark for classifi- cation only reaches an accuracy of 0.8 which means that the two data sets do not fully agree on melt areas. For our project, this means that we have to trust that the optical data set is superior to the microwave data set in its accuracy. If this assump- tion holds true, we can use the optical data as labels and we can improve melt map precision. But we need to investigate this matter further by checking the overlap of the two data sets closer, especially around the melt boundary.