U-SEG: Uncertainty in SEGmentation -- A systematic multi-variable exploration

TL;DR

This paper systematically investigates how various factors affect uncertainty estimation in segmentation tasks, revealing insights into their practical utility and limitations across different scenarios.

Contribution

It provides a comprehensive framework and large-scale analysis of uncertainty estimation methods across multiple datasets, architectures, and tasks in segmentation.

Findings

Panoptic segmentation is more challenging and less generalizable.

Time series samples can be useful but often are not cost-effective.

Sample diversity improves calibration but does not outperform simpler methods.

Abstract

In this study, we explore in depth a few under-studied topics at the intersection of uncertainty estimation and segmentation. Prior work has shown that the quality of uncertainty estimates can be very sensitive to a range of variables. As one of the main uses of uncertainty estimation is to help identify and deal with prediction errors in practical scenarios, any factors that affect this must be clearly identified. For example, do more challenging domains or different datasets and architectures result in worse performance when using uncertainty estimates? Can prior frames in a video sequence in fact provide useful uncertainty estimates comparable to other approaches? Is it possible to combine uncertainty estimation approaches, taking advantage of sample diversity, to get better estimates? Finally, when might it make sense to use an ensemble-based uncertainty estimate over a deterministic network? We address these questions by creating a framework for and executing a large scale study across many variables such as datasets, backbones, and downstream tasks, for both semantic and panoptic segmentation. We find that a) the more challenging task of panoptic segmentation usually results in worse performance while high performance variance between datasets and backbones indicates that generalization is not guaranteed, b) time series samples can be useful for specific configurations, but in many cases are not worth the cost, c) sample diversity shows the most promise in the downstream task of calibration, but otherwise fails to beat simpler alternatives, d) a deterministic approach is adequate for some downstream tasks, but ensembles allow for significant improvements if the right conditions can be achieved in deployment.