What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?

–Neural Information Processing Systems 

There are two major types of uncertainty one can model.