VOS: Learning What You Don't Know by Virtual Outlier Synthesis

By Xuefeng Du, Zhaoning Wang, Mu Cai, Yixuan Li. International Conference on Learning Representations, 2022

447 citations recorded by the source corpus.

Out-of-distribution (OOD) detection has received much attention lately due to its importance in the safe deployment of neural networks. One of the key challenges is that models lack supervision signals from unknown data, and as a result, can produce overconfident predictions on OOD data. Previous approaches rely on real outlier datasets for model regularization, which can be costly and sometimes infeasible to obtain in practice. In this paper, we present VOS, a novel framework for OOD detection by adaptively synthesizing virtual outliers that can meaningfully regularize the model's decision boundary during training. Specifically, VOS samples virtual outliers from the low-likelihood region of the class-conditional distribution estimated in the feature space. Alongside, we introduce a novel unknown-aware training objective, which contrastively shapes the uncertainty space between the ID data and synthesized outlier data. VOS achieves competitive performance on both object detection and image classification models, reducing the FPR95 by up to 9.36% compared to the previous best method on object detectors. Code is available at https://github.com/deeplearning-wisc/vos.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/f0e6aed660c541544063c0608326c6f19ba48f6d

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Ziwei Liu
2. Yixuan Li
3. Kaiyang Zhou
4. Jingkang Yang
5. Wayne Zhang
6. Haoqi Wang
7. Bo Li
8. Yiyou Sun

Citations in the corpus

• Generalized Out-of-Distribution Detection: A Survey (2021, unknown): “In particular, recent works start looking more into object-level OOD detection [146], [147], which can be useful for safety-critical”
• OpenOOD: Benchmarking Generalized Out-of-Distribution Detection (2022, background): “Future Work In the future, apart from maintaining the codebase, it is promising to extend our benchmark towards more robust OOD detectors [82] and object-level OOD detection and generalization [69, 83, 84, 85, 86, 87], which provides finer-grained visual guidance in safety-critical applications, suc”
• Mitigating Neural Network Overconfidence with Logit Normalization (2022, background): “For example, models are encouraged to give predictions with uniform distribution (Lee et al., 2017; Hendrycks et al., 2019) or higher energies (Liu et al., 2020; Du et al., 2022b; Ming et al., 2022a; Du et al., 2022a; Katz-Samuels et al., 2022) for outliers.”