CODEs: Chamfer Out-of-Distribution Examples against Overconfidence Issue

TL;DR

This paper introduces Chamfer OOD examples (CODEs), generated close to in-distribution data, to effectively reduce overconfidence on OOD samples and improve detection without extra data or accuracy loss.

Contribution

The paper proposes a novel method to generate OOD examples near in-distribution data using Chamfer GANs, enhancing OOD detection and confidence calibration.

Findings

Training with CODEs reduces OOD overconfidence significantly.

The method outperforms existing state-of-the-art approaches.

CODEs improve both OOD detection and classification accuracy.

Abstract

Overconfident predictions on out-of-distribution (OOD) samples is a thorny issue for deep neural networks. The key to resolve the OOD overconfidence issue inherently is to build a subset of OOD samples and then suppress predictions on them. This paper proposes the Chamfer OOD examples (CODEs), whose distribution is close to that of in-distribution samples, and thus could be utilized to alleviate the OOD overconfidence issue effectively by suppressing predictions on them. To obtain CODEs, we first generate seed OOD examples via slicing&splicing operations on in-distribution samples from different categories, and then feed them to the Chamfer generative adversarial network for distribution transformation, without accessing to any extra data. Training with suppressing predictions on CODEs is validated to alleviate the OOD overconfidence issue largely without hurting classification accuracy, and outperform the state-of-the-art methods. Besides, we demonstrate CODEs are useful for improving OOD detection and classification.