Is segmentation uncertainty useful?

TL;DR

This paper evaluates how different probabilistic segmentation models encode uncertainty and their effectiveness in assessing segmentation quality and active learning, finding that uncertainty correlates with error but is not useful for active learning.

Contribution

It compares four probabilistic segmentation models, analyzing their uncertainty modeling and performance in quality assessment and active learning tasks.

Findings

Uncertainty correlates positively with segmentation error.

Uncertainty is not effective for active learning.

Different models capture various aspects of segmentation ambiguity.

Abstract

Probabilistic image segmentation encodes varying prediction confidence and inherent ambiguity in the segmentation problem. While different probabilistic segmentation models are designed to capture different aspects of segmentation uncertainty and ambiguity, these modelling differences are rarely discussed in the context of applications of uncertainty. We consider two common use cases of segmentation uncertainty, namely assessment of segmentation quality and active learning. We consider four established strategies for probabilistic segmentation, discuss their modelling capabilities, and investigate their performance in these two tasks. We find that for all models and both tasks, returned uncertainty correlates positively with segmentation error, but does not prove to be useful for active learning.