DynProto: Dynamic Prototype Evolution for Out-of-Distribution Detection

TL;DR

DynProto introduces a dynamic prototype learning method during testing that improves out-of-distribution detection by clustering and refining OOD patterns using only in-distribution data.

Contribution

It proposes a novel, architecture-agnostic framework that dynamically learns OOD prototypes during testing, addressing limitations of previous methods relying on predefined OOD labels.

Findings

DynProto reduces FPR95 by 11.60% on ImageNet OOD benchmark.

It improves AUROC by 4.70% on the same benchmark.

The method outperforms prior approaches across multiple benchmarks.

Abstract

Recent studies show that using potential out-of-distribution (OOD) labels from large corpora as auxiliary information can improve OOD detection in vision-language models (VLMs). However, these methods often fail when real-world OOD samples fall outside the predefined OOD label set. To address this limitation, we propose DynProto, a novel approach that learns OOD prototypes dynamically during testing using only in-distribution (ID) information. DynProto is inspired by a key observation: OOD samples predicted as the same ID class tend to cluster in the feature space. With this insight, we leverage easy-to-detect OOD samples as ``anchors'' to find their harder-to-detect, similar counterparts. To this end, DynProto introduces two modules: \textbf{Coarse OOD Pattern Capturing Module} caches OOD patterns that are easily confused with each ID class during testing, and \textbf{Fine-grained OOD Pattern Refinement Module} subsequently clusters these patterns within each cache and aggregates them into representative OOD prototypes. By measuring similarity to ID and dynamic OOD prototypes, DynProto enables accurate OOD detection. DynProto significantly outperforms prior methods across multiple benchmarks. On ImageNet OOD benchmark, DynProto reduces FPR95 by 11.60\% and improves AUROC by 4.70\%. Moreover, the framework is architecture-agnostic and can be integrated into various backbones.