DAVIS: OOD Detection via Dominant Activations and Variance for Increased Separation

2026-01-30

AI Summary

This work addresses a key limitation in existing out-of-distribution (OOD) detection methods, which rely solely on global average pooling (GAP) and thus exploit only the mean activation values while neglecting higher-order channel-wise statistics. To overcome this, the authors propose DAVIS, a general-purpose post-hoc method that, for the first time, incorporates channel-wise variance and maximum activation prior to GAP to enrich feature representations and enhance separability between in-distribution and OOD samples. Notably, DAVIS requires no modification to the underlying model architecture and is compatible with mainstream backbones such as ResNet, DenseNet, and EfficientNet. Extensive experiments demonstrate substantial improvements, reducing FPR95 by 48.26%, 38.13%, and 26.83% on CIFAR-10, CIFAR-100, and ImageNet-1k benchmarks, respectively, thereby setting new state-of-the-art results across multiple OOD detection tasks.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Other Foundations of Computer Vision
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Detecting out-of-distribution (OOD) inputs is a critical safeguard for deploying machine learning models in the real world. However, most post-hoc detection methods operate on penultimate feature representations derived from global average pooling (GAP) -- a lossy operation that discards valuable distributional statistics from activation maps prior to global average pooling. We contend that these overlooked statistics, particularly channel-wise variance and dominant (maximum) activations, are highly discriminative for OOD detection. We introduce DAVIS, a simple and broadly applicable post-hoc technique that enriches feature vectors by incorporating these crucial statistics, directly addressing the information loss from GAP. Extensive evaluations show DAVIS sets a new benchmark across diverse architectures, including ResNet, DenseNet, and EfficientNet. It achieves significant reductions in the false positive rate (FPR95), with improvements of 48.26\% on CIFAR-10 using ResNet-18, 38.13\% on CIFAR-100 using ResNet-34, and 26.83\% on ImageNet-1k benchmarks using MobileNet-v2. Our analysis reveals the underlying mechanism for this improvement, providing a principled basis for moving beyond the mean in OOD detection.

Problem

• out-of-distribution detection
• OOD detection
• distributional statistics
• feature representation
• model robustness

Innovation

• out-of-distribution detection
• channel-wise variance
• dominant activations
• global average pooling
• feature enrichment