Distributional Gaussian Process Layers for Outlier Detection in Image Segmentation

TL;DR

This paper introduces a novel Bayesian convolutional Gaussian Process layer that effectively propagates uncertainty in image segmentation, achieving near state-of-the-art performance and superior out-of-distribution detection, especially in medical imaging.

Contribution

It presents a parameter-efficient hierarchical Gaussian Process layer using Wasserstein-2 space, enabling reliable uncertainty propagation and out-of-distribution detection in segmentation tasks.

Findings

Achieves performance close to deterministic U-Net in brain tissue segmentation.

Outperforms previous Bayesian methods in out-of-distribution detection.

Provides reliable uncertainty estimates for pathology detection.

Abstract

We propose a parameter efficient Bayesian layer for hierarchical convolutional Gaussian Processes that incorporates Gaussian Processes operating in Wasserstein-2 space to reliably propagate uncertainty. This directly replaces convolving Gaussian Processes with a distance-preserving affine operator on distributions. Our experiments on brain tissue-segmentation show that the resulting architecture approaches the performance of well-established deterministic segmentation algorithms (U-Net), which has never been achieved with previous hierarchical Gaussian Processes. Moreover, by applying the same segmentation model to out-of-distribution data (i.e., images with pathology such as brain tumors), we show that our uncertainty estimates result in out-of-distribution detection that outperforms the capabilities of previous Bayesian networks and reconstruction-based approaches that learn normative distributions.