Lune

ICLR2022Top-tier venue

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

Xuefeng Du, Zhaoning Wang, Mu Cai, Yixuan Li

Abstract

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.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

143 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learning cites Delving into Out-of-Distribution Detection with Vision-Language Representations
  • LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learning cites Energy-based Out-of-distribution Detection
  • LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learning cites On the Importance of Gradients for Detecting Distributional Shifts in the Wild
  • LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learning cites ReAct: Out-of-distribution Detection With Rectified Activations
  • LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learning cites Open-Set Recognition: A Good Closed-Set Classifier is All You Need
  • LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learning cites SIREN: Shaping Representations for Detecting Out-of-Distribution Objects
  • Delving into Out-of-Distribution Detection with Vision-Language Representations cites Energy-based Out-of-distribution Detection
  • Delving into Out-of-Distribution Detection with Vision-Language Representations cites On the Importance of Gradients for Detecting Distributional Shifts in the Wild
  • Delving into Out-of-Distribution Detection with Vision-Language Representations cites ReAct: Out-of-distribution Detection With Rectified Activations
  • Delving into Out-of-Distribution Detection with Vision-Language Representations cites Detecting Out-of-Distribution Examples with Gram Matrices
  • Delving into Out-of-Distribution Detection with Vision-Language Representations cites Open-Set Recognition: A Good Closed-Set Classifier is All You Need
  • Delving into Out-of-Distribution Detection with Vision-Language Representations cites SIREN: Shaping Representations for Detecting Out-of-Distribution Objects
  • Delving into Out-of-Distribution Detection with Vision-Language Representations cites CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances
  • On the Importance of Gradients for Detecting Distributional Shifts in the Wild cites Energy-based Out-of-distribution Detection
  • ReAct: Out-of-distribution Detection With Rectified Activations cites Energy-based Out-of-distribution Detection
  • ReAct: Out-of-distribution Detection With Rectified Activations cites On the Importance of Gradients for Detecting Distributional Shifts in the Wild
  • Open-Set Recognition: A Good Closed-Set Classifier is All You Need cites Energy-based Out-of-distribution Detection
  • Open-Set Recognition: A Good Closed-Set Classifier is All You Need cites CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances
  • SIREN: Shaping Representations for Detecting Out-of-Distribution Objects cites Energy-based Out-of-distribution Detection
  • SIREN: Shaping Representations for Detecting Out-of-Distribution Objects cites ReAct: Out-of-distribution Detection With Rectified Activations
  • SIREN: Shaping Representations for Detecting Out-of-Distribution Objects cites Detecting Out-of-Distribution Examples with Gram Matrices
  • SIREN: Shaping Representations for Detecting Out-of-Distribution Objects cites CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances
  • Dream the Impossible: Outlier Imagination with Diffusion Models cites Energy-based Out-of-distribution Detection
  • Dream the Impossible: Outlier Imagination with Diffusion Models cites On the Importance of Gradients for Detecting Distributional Shifts in the Wild
  • Dream the Impossible: Outlier Imagination with Diffusion Models cites ReAct: Out-of-distribution Detection With Rectified Activations
  • Dream the Impossible: Outlier Imagination with Diffusion Models cites SIREN: Shaping Representations for Detecting Out-of-Distribution Objects
  • Dream the Impossible: Outlier Imagination with Diffusion Models cites CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers143

  1. Open-Set Recognition: A Good Closed-Set Classifier is All You NeedSagar Vaze, Kai Han, Andrea Vedaldi, Andrew ZissermanICLR 2022 · 594 citations
  2. Delving into Out-of-Distribution Detection with Vision-Language RepresentationsYifei Ming, Ziyang Cai, Jiuxiang Gu, Yiyou Sun et al.NeurIPS 2022 · 308 citations
  3. LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt LearningAtsuyuki Miyai, Qing Yu, Go Irie, Kiyoharu AizawaNeurIPS 2023 · 174 citations
  4. Dream the Impossible: Outlier Imagination with Diffusion ModelsXuefeng Du, Yiyou Sun, Jerry Zhu, Yixuan LiNeurIPS 2023 · 114 citations
  5. SIREN: Shaping Representations for Detecting Out-of-Distribution ObjectsXuefeng Du, Gabriel Gozum, Yifei Ming, Yixuan LiNeurIPS 2022 · 105 citations

Builds on15

  1. Energy-based Out-of-distribution DetectionWeitang Liu, Xiaoyun Wang, John D. Owens, Yixuan LiNeurIPS 2020 · 2,213 citations
  2. CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted InstancesJihoon Tack, Sangwoo Mo, Jongheon Jeong, Jinwoo ShinNeurIPS 2020 · 755 citations
  3. ReAct: Out-of-distribution Detection With Rectified ActivationsYiyou Sun, Chuan Guo, Yixuan LiNeurIPS 2021 · 733 citations
  4. On the Importance of Gradients for Detecting Distributional Shifts in the WildRui Huang, Andrew Geng, Yixuan LiNeurIPS 2021 · 515 citations
  5. Detecting Out-of-Distribution Examples with Gram MatricesChandramouli Shama Sastry, Sageev OoreICML 2020 · 275 citations

Related papers

  1. Out-of-Distribution Detection with Prototypical Outlier ProxyMingrong Gong, Chaoqi Chen, Qingqiang Sun, Yue Wang et al.AAAI 2025 · 8 citations
  2. Non-parametric Outlier SynthesisLeitian Tao, Xuefeng Du, Jerry Zhu, Yixuan LiICLR 2023 · 6 citations
  3. Geometrically Constrained Outlier SynthesisDaniil Karzanov, Marcin DetynieckiICML 2026 · 1 citation
  4. Density-driven Regularization for Out-of-distribution DetectionWenjian Huang, Hao Wang, Jiahao Xia, Chengyan Wang et al.NeurIPS 2022 · 17 citations
  5. Learning Latent Concepts for Detecting Out-of-Distribution ObjectsTing Peng, Junhao Dong, Yew-Soon OngCVPR 2026
Dusk over the sea between two cliffs drawn in fine vertical lines