3 images taken by NASA's Multi-angle Imaging SpectroRadiometer imagery, a satelitte that has many different cameras at various incidence angles collecting electromagnetic radiation at different wavelenghts. Normally, the ability for machine learning methods to detect cloud from land based on these readings is very easy. But near the polar caps, because both land and clounds share similiar reflective properties, usual methods work very poorly.
In this lab we build and compare models to distinguish between clouds and snow using random forest, logistic regression, LDA and QDA. The detailed report, comparisons with performance graphs can be found in writeupLab4.
This lab is reproducible, and instructions for doing so can be found in how_to_reproduce.