Out-of-distribution detection based on subspace projection of high-dimensional features output by the last convolutional layer

TL;DR

This paper introduces a novel OOD detection method that projects high-dimensional features into subspaces, modifies network layers to enhance separation, and achieves state-of-the-art results without requiring OOD data during training.

Contribution

The paper proposes a new OOD detection approach based on subspace projection of features and layer modifications, improving detection performance without extra OOD data.

Findings

Achieves state-of-the-art OOD detection accuracy on benchmark datasets.

Requires only training the classification model, no input pre-processing or OOD tuning.

Effective in distinguishing in-distribution and OOD data through feature subspace analysis.

Abstract

Out-of-distribution (OOD) detection, crucial for reliable pattern classification, discerns whether a sample originates outside the training distribution. This paper concentrates on the high-dimensional features output by the final convolutional layer, which contain rich image features. Our key idea is to project these high-dimensional features into two specific feature subspaces, leveraging the dimensionality reduction capacity of the network's linear layers, trained with Predefined Evenly-Distribution Class Centroids (PEDCC)-Loss. This involves calculating the cosines of three projection angles and the norm values of features, thereby identifying distinctive information for in-distribution (ID) and OOD data, which assists in OOD detection. Building upon this, we have modified the batch normalization (BN) and ReLU layer preceding the fully connected layer, diminishing their impact on the output feature distributions and thereby widening the distribution gap between ID and OOD data features. Our method requires only the training of the classification network model, eschewing any need for input pre-processing or specific OOD data pre-tuning. Extensive experiments on several benchmark datasets demonstrates that our approach delivers state-of-the-art performance. Our code is available at https://github.com/Hewell0/ProjOOD.