Rethinking Out-of-Distribution Detection on Imbalanced Data Distribution

TL;DR

This paper introduces a statistical framework called ImOOD to address the challenges of out-of-distribution detection in imbalanced data, proposing a regularization method that improves performance on multiple benchmarks.

Contribution

It formulates OOD detection on imbalanced data, reveals class-aware bias issues, and proposes a training regularization technique to enhance detection accuracy.

Findings

Improved OOD detection on CIFAR10-LT, CIFAR100-LT, and ImageNet-LT.

Identified class-aware bias as a key factor in performance gaps.

Proposed method outperforms several state-of-the-art approaches.

Abstract

Detecting and rejecting unknown out-of-distribution (OOD) samples is critical for deployed neural networks to void unreliable predictions. In real-world scenarios, however, the efficacy of existing OOD detection methods is often impeded by the inherent imbalance of in-distribution (ID) data, which causes significant performance decline. Through statistical observations, we have identified two common challenges faced by different OOD detectors: misidentifying tail class ID samples as OOD, while erroneously predicting OOD samples as head class from ID. To explain this phenomenon, we introduce a generalized statistical framework, termed ImOOD, to formulate the OOD detection problem on imbalanced data distribution. Consequently, the theoretical analysis reveals that there exists a class-aware bias item between balanced and imbalanced OOD detection, which contributes to the performance gap. Building upon this finding, we present a unified training-time regularization technique to mitigate the bias and boost imbalanced OOD detectors across architecture designs. Our theoretically grounded method translates into consistent improvements on the representative CIFAR10-LT, CIFAR100-LT, and ImageNet-LT benchmarks against several state-of-the-art OOD detection approaches. Code is available at https://github.com/alibaba/imood.