Hybrid Energy Based Model in the Feature Space for Out-of-Distribution Detection

TL;DR

The paper presents HEAT, a hybrid energy-based model that improves out-of-distribution detection by accurately estimating in-distribution sample density in the feature space, achieving state-of-the-art results on multiple benchmarks.

Contribution

Introduction of HEAT, a novel post-hoc OOD detection method using hybrid energy-based models for density estimation in feature space, enhancing robustness and accuracy.

Findings

HEAT achieves state-of-the-art OOD detection on CIFAR-10 and CIFAR-100.

HEAT outperforms existing density estimators like GMM.

Extensive experiments validate the effectiveness of the hybrid energy-based approach.

Abstract

Out-of-distribution (OOD) detection is a critical requirement for the deployment of deep neural networks. This paper introduces the HEAT model, a new post-hoc OOD detection method estimating the density of in-distribution (ID) samples using hybrid energy-based models (EBM) in the feature space of a pre-trained backbone. HEAT complements prior density estimators of the ID density, e.g. parametric models like the Gaussian Mixture Model (GMM), to provide an accurate yet robust density estimation. A second contribution is to leverage the EBM framework to provide a unified density estimation and to compose several energy terms. Extensive experiments demonstrate the significance of the two contributions. HEAT sets new state-of-the-art OOD detection results on the CIFAR-10 / CIFAR-100 benchmark as well as on the large-scale Imagenet benchmark. The code is available at: https://github.com/MarcLafon/heatood.