Deep neural networks are susceptible to generating overconfident yet erroneous predictions when presented with data beyond known concepts. This challenge underscores the importance of detecting out-of-distribution (OOD) samples in the open world. In this work, we propose a novel feature-space OOD detection score that jointly reasons with both class-specific and class-agnostic information. Specifically, our approach utilizes Whitened Linear Discriminative Analysis to project features into two subspaces - the discriminative and residual subspaces - in which the ID classes are maximally separated and closely clustered, respectively. The OOD score is then determined by combining the deviation from the input data to the ID distribution in both subspaces. The efficacy of our method, named WDiscOOD, is verified on the large-scale ImageNet-1k benchmark, with six OOD datasets that covers a variety of distribution shifts. WDiscOOD demonstrates superior performance on deep classifiers with diverse backbone architectures, including CNN and vision transformer. Furthermore, we also show that our method can more effectively detect novel concepts in representation space trained with contrastive objectives, including supervised contrastive loss and multi-modality contrastive loss.

翻译：深度神经网络在遇到已知概念之外的数据时，容易产生过度自信但错误的预测。这一挑战凸显了在开放世界中检测分布外（OOD）样本的重要性。本文提出了一种新颖的特征空间OOD检测分数，该分数联合推理了类别特定与类别无关信息。具体而言，我们的方法利用白化线性判别分析将特征投影到两个子空间——判别子空间和残差子空间——其中ID类别分别实现最大程度分离和紧密聚类。随后，通过结合输入数据在两个子空间中偏离ID分布的程度来确定OOD分数。所提出的方法WDiscOOD在大规模ImageNet-1k基准测试以及覆盖多种分布偏移的六个OOD数据集上得到了验证。WDiscOOD在具有多种骨干架构（包括CNN和视觉Transformer）的深度分类器上展现了卓越性能。此外，我们还表明，该方法能更有效地检测通过对比学习目标（包括监督对比损失和多模态对比损失）训练得到的表示空间中的新概念。