Robustness to Spurious Correlations Improves Semantic Out-of-Distribution Detection

TL;DR

This paper introduces nuisance-aware methods for out-of-distribution detection that are robust to spurious correlations, significantly improving detection of shared-nuisance OOD inputs by training classifiers to be independent of nuisance factors.

Contribution

The paper proposes a novel nuisance-aware OOD detection approach using Nuisance-Randomized Distillation (NuRD) to improve robustness against spurious correlations in OOD detection.

Findings

Nuisance-aware detection outperforms traditional methods on shared-nuisance OOD inputs.

NuRD-based classifiers achieve better independence from nuisance factors.

The approach remains effective even when domain generalization methods fail.

Abstract

Methods which utilize the outputs or feature representations of predictive models have emerged as promising approaches for out-of-distribution (OOD) detection of image inputs. However, these methods struggle to detect OOD inputs that share nuisance values (e.g. background) with in-distribution inputs. The detection of shared-nuisance out-of-distribution (SN-OOD) inputs is particularly relevant in real-world applications, as anomalies and in-distribution inputs tend to be captured in the same settings during deployment. In this work, we provide a possible explanation for SN-OOD detection failures and propose nuisance-aware OOD detection to address them. Nuisance-aware OOD detection substitutes a classifier trained via empirical risk minimization and cross-entropy loss with one that 1. is trained under a distribution where the nuisance-label relationship is broken and 2. yields representations that are independent of the nuisance under this distribution, both marginally and conditioned on the label. We can train a classifier to achieve these objectives using Nuisance-Randomized Distillation (NuRD), an algorithm developed for OOD generalization under spurious correlations. Output- and feature-based nuisance-aware OOD detection perform substantially better than their original counterparts, succeeding even when detection based on domain generalization algorithms fails to improve performance.