利用特征图原始统计量提升分布外检测效果
Catalyst: Out-of-Distribution Detection via Elastic Scaling
- 从池化前特征图中提取通道统计量动态生成缩放因子
- 在CIFAR-10/100和ImageNet上平均误报率降低超22%
- 兼容主流检测方法，适合部署安全要求高的场景
分布外（OOD）检测对深度神经网络的安全部署至关重要。现有先进后处理方法通常基于全局平均池化（GAP）后的输出logits或倒数第二层特征向量计算OOD得分。本文指出，这种仅依赖logits或特征向量的做法忽略了被GAP丢弃的丰富互补信号：池化前特征图的原始通道级统计信息。为此，我们提出Catalyst框架，利用这些未被充分挖掘的信号。Catalyst通过实时计算输入相关的缩放因子γ（基于均值、标准差、最大激活等原始统计量），并将该γ与基线得分相乘进行弹性缩放，从而拉大ID与OOD分布的距离。实验表明，Catalyst具有强泛化能力：可无缝集成于logit-based方法（如Energy、ReAct、SCALE），并显著提升KNN等距离型检测器性能。结果在CIFAR-10（ResNet-18）上将平均误报率降低32.87%，CIFAR-100（ResNet-18）降低27.94%，ImageNet（ResNet-50）降低22.25%。这揭示了预池化统计量的巨大潜力，并证明Catalyst与现有方法具有互补性。代码已开源：https://github.com/bingabid/Catalyst
原文摘要 · Abstract (English)
Out-of-distribution (OOD) detection is critical for the safe deployment of deep neural networks. State-of-the-art post-hoc methods typically derive OOD scores from the output logits or penultimate feature vector obtained via global average pooling (GAP). We contend that this exclusive reliance on the logit or feature vector discards a rich, complementary signal: the raw channel-wise statistics of the pre-pooling feature map lost in GAP. In this paper, we introduce Catalyst, a post-hoc framework that exploits these under-explored signals. Catalyst computes an input-dependent scaling factor ($γ$) on-the-fly from these raw statistics (e.g., mean, standard deviation, and maximum activation). This $γ$ is then fused with the existing baseline score, multiplicatively modulating it -- an $\textit{elastic scaling}$ -- to push the ID and OOD distributions further apart. We demonstrate Catalyst is a generalizable framework: it seamlessly integrates with logit-based methods (e.g., Energy, ReAct, SCALE) and also provides a significant boost to distance-based detectors like KNN. As a result, Catalyst achieves substantial and consistent performance gains, reducing the average False Positive Rate by 32.87 on CIFAR-10 (ResNet-18), 27.94% on CIFAR-100 (ResNet-18), and 22.25% on ImageNet (ResNet-50). Our results highlight the untapped potential of pre-pooling statistics and demonstrate that Catalyst is complementary to existing OOD detection approaches. Our code is available here: https://github.com/bingabid/Catalyst
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。