Probing Contextual Diversity for Dense Out-of-Distribution Detection

TL;DR

This paper introduces MOoSe, a method that leverages multi-scale context in segmentation models to improve dense out-of-distribution detection and uncertainty estimation.

Contribution

The paper proposes MOoSe, a novel approach that effectively utilizes multi-scale contextual information to enhance OoD detection in semantic segmentation tasks.

Findings

Multi-scale aggregation improves OoD detection accuracy.

MOoSe enhances uncertainty estimation in segmentation models.

Simple multi-scale strategies yield consistent benefits.

Abstract

Detection of out-of-distribution (OoD) samples in the context of image classification has recently become an area of interest and active study, along with the topic of uncertainty estimation, to which it is closely related. In this paper we explore the task of OoD segmentation, which has been studied less than its classification counterpart and presents additional challenges. Segmentation is a dense prediction task for which the model's outcome for each pixel depends on its surroundings. The receptive field and the reliance on context play a role for distinguishing different classes and, correspondingly, for spotting OoD entities. We introduce MOoSe, an efficient strategy to leverage the various levels of context represented within semantic segmentation models and show that even a simple aggregation of multi-scale representations has consistently positive effects on OoD detection and uncertainty estimation.