Feature-Space Semantic Invariance: Enhanced OOD Detection for Open-Set Domain Generalization

TL;DR

A unified framework for open-set domain generalization is proposed by introducing Feature-space Semantic Invariance (FSI), which maintains semantic consistency across different domains within the feature space, enabling more accurate detection of OOD instances in unseen domains.

Abstract

Open-set domain generalization addresses a real-world challenge: training a model to generalize across unseen domains (domain generalization) while also detecting samples from unknown classes not encountered during training (open-set recognition). However, most existing approaches tackle these issues separately, limiting their practical applicability. To overcome this limitation, we propose a unified framework for open-set domain generalization by introducing Feature-space Semantic Invariance (FSI). FSI maintains semantic consistency across different domains within the feature space, enabling more accurate detection of OOD instances in unseen domains. Additionally, we adopt a generative model to produce synthetic data with novel domain styles or class labels, enhancing model robustness. Initial experiments show that our method improves AUROC by 9.1% to 18.9% on ColoredMNIST, while also significantly increasing in-distribution classification accuracy.

Figures (2)

• Figure 1: An illustration of the proposed framework. The framework samples instances from different domains, each characterized by unique variations (e.g., colors), aiming to learn a domain-invariant feature extractor that can be combined with state-of-the-art semantic OOD detectors to effectively address both domain generalization and open-set recognition challenges.
• Figure 2: Examples of synthetic OODs generated by blending two ID samples.

Theorems & Definitions (1)

• Definition 1: Feature-space Semantic Invariance (FSI)