OpenMix: Exploring Outlier Samples for Misclassification Detection
About
Reliable confidence estimation for deep neural classifiers is a challenging yet fundamental requirement in high-stakes applications. Unfortunately, modern deep neural networks are often overconfident for their erroneous predictions. In this work, we exploit the easily available outlier samples, i.e., unlabeled samples coming from non-target classes, for helping detect misclassification errors. Particularly, we find that the well-known Outlier Exposure, which is powerful in detecting out-of-distribution (OOD) samples from unknown classes, does not provide any gain in identifying misclassification errors. Based on these observations, we propose a novel method called OpenMix, which incorporates open-world knowledge by learning to reject uncertain pseudo-samples generated via outlier transformation. OpenMix significantly improves confidence reliability under various scenarios, establishing a strong and unified framework for detecting both misclassified samples from known classes and OOD samples from unknown classes. The code is publicly available at https://github.com/Impression2805/OpenMix.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| OOD Detection | CIFAR-100 standard (test) | AUROC (%)84.88 | 94 | |
| Out-of-Distribution Detection | CIFAR100 | AURC342.2 | 39 | |
| Failure Detection | CIFAR100 vs. SVHN | AURC Score406.8 | 39 | |
| Failure Detection | CIFAR100 (test) | AURC85.66 | 39 | |
| Misclassification Detection | CIFAR-10 | AUROC94.81 | 28 | |
| Misclassification Detection | CIFAR-100 | AURC73.84 | 27 | |
| Out-of-Distribution Detection | CIFAR-10 (ID) vs 6 OOD datasets (Textures, SVHN, Place365, LSUN-C, LSUN-R, iSUN) (test) | FPR@9516.86 | 24 | |
| Misclassification Detection | CIFAR-10-C 1.0 (test) | AUROC90.38 | 9 | |
| Misclassification Detection | CIFAR-100-C 1.0 (test) | AUROC84.05 | 9 |