Generalized ODIN: Detecting Out-of-distribution Image without Learning from Out-of-distribution Data

2020/02/26 by Yen-Chang Hsu, Yilin Shen, Hsu, Yen-Chang +5 · 48 citations
Computer Science · Engineering · Mathematics · #A priori and a posteriori #Advanced Neural Network Applications #Adversarial Robustness in Machine Learning #Anomaly Detection Techniques and Applications #Artificial intelligence #Artificial neural network #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Data mining #Data set #Distribution (mathematics) #FOS: Computer and information sciences #FOS: Electrical engineering #Image (mathematics) #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Machine learning #Mathematics #Pattern recognition (psychology) #Scale (ratio) #Selection (genetic algorithm) #Set (abstract data type) #Training set #cs.CV #cs.LG #eess.IV #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2002.11297

published in arXiv (Cornell University) (Cornell University) · CVPR 2020

openalex publication_date 2020/02/26 · openalex created_date 2020/03/06 · arxiv created 2020/03/31 · arxiv updated 2020/04/02 · openalex updated_date 2026/08/05

Abstract

Deep neural networks have attained remarkable performance when applied to data that comes from the same distribution as that of the training set, but can significantly degrade otherwise. Therefore, detecting whether an example is out-of-distribution (OoD) is crucial to enable a system that can reject such samples or alert users. Recent works have made significant progress on OoD benchmarks consisting of small image datasets. However, many recent methods based on neural networks rely on training or tuning with both in-distribution and out-of-distribution data. The latter is generally hard to define a-priori, and its selection can easily bias the learning. We base our work on a popular method ODIN, proposing two strategies for freeing it from the needs of tuning with OoD data, while improving its OoD detection performance. We specifically propose to decompose confidence scoring as well as a modified input pre-processing method. We show that both of these significantly help in detection performance. Our further analysis on a larger scale image dataset shows that the two types of distribution shifts, specifically semantic shift and non-semantic shift, present a significant difference in the difficulty of the problem, providing an analysis of when ODIN-like strategies do or do not work.

Citations

• Visual Domain Adaptation: A survey of recent advances
• Wide Residual Networks
• A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks
• Reducing Network Agnostophobia
• Deep Anomaly Detection with Outlier Exposure
• Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
• PixelCNN++: Improving the PixelCNN with Discretized Logistic Mixture Likelihood and Other Modifications
• Learning Multiple Layers of Features from Tiny Images

Cited by

• Training-Free Policy Violation Detection via Activation-Space Whitening in LLMs
• RankOOD -- Class Ranking-based Out-of-Distribution Detection
• SG-OIF: A Stability-Guided Online Influence Framework for Reliable Vision Data
• BootOOD: Self-Supervised Out-of-Distribution Detection via Synthetic Sample Exposure under Neural Collapse
• Out-of-Distribution Detection for Safety Assurance of AI and Autonomous Systems
• Revisiting Logit Distributions for Reliable Out-of-Distribution Detection
• Beyond Binary Out-of-Distribution Detection: Characterizing Distributional Shifts with Multi-Statistic Diffusion Trajectories
• GOOD: Training-Free Guided Diffusion Sampling for Out-of-Distribution Detection
• Local Background Features Matter in Out-of-Distribution Detection
• Equipping Vision Foundation Model with Mixture of Experts for Out-of-Distribution Detection
• Human Texts Are Outliers: Detecting LLM-generated Texts via Out-of-distribution Detection
• Long-Tailed Out-of-Distribution Detection with Refined Separate Class Learning
• Dynamic Aware: Adaptive Multi-Mode Out-of-Distribution Detection for Trajectory Prediction in Autonomous Vehicles
• An Empirical Analysis of VLM-based OOD Detection: Mechanisms, Advantages, and Sensitivity
• Out of Distribution Detection in Self-adaptive Robots with AI-powered Digital Twins
• Sharpness-Aware Geometric Defense for Robust Out-Of-Distribution Detection
• Optimizing Latent Dimension Allocation in Hierarchical VAEs: Balancing Attenuation and Information Retention for OOD Detection
• ANTS: Adaptive Negative Textual Space Shaping for OOD Detection via Test-Time MLLM Understanding and Reasoning
• What Can We Learn from Harry Potter? An Exploratory Study of Visual Representation Learning from Atypical Videos
• SNAP-UQ: Self-supervised Next-Activation Prediction for Single-Pass Uncertainty in TinyML
• OpenHAIV: A Framework Towards Practical Open-World Learning
• COSTARR: Consolidated Open Set Technique with Attenuation for Robust Recognition
• A Simple and Effective Method for Uncertainty Quantification and OOD Detection
• BOOD: Boundary-based Out-Of-Distribution Data Generation
• Out-of-Distribution Detection in Medical Imaging via Diffusion Trajectories
• Label tree semantic losses for rich multi-class medical image segmentation
• Handling Out-of-Distribution Data: A Survey
• Out-of-distribution data supervision towards biomedical semantic segmentation
• Feature Bank Enhancement for Distance-based Out-of-Distribution Detection
• FA: Forced Prompt Learning of Vision-Language Models for Out-of-Distribution Detection
• Tree-based Semantic Losses: Application to Sparsely-supervised Large Multi-class Hyperspectral Segmentation
• Adaptive Multi-prompt Contrastive Network for Few-shot Out-of-distribution Detection
• Balanced Hyperbolic Embeddings Are Natural Out-of-Distribution Detectors
• TRUST: Test-time Resource Utilization for Superior Trustworthiness
• Enhancing Abnormality Identification: Robust Out-of-Distribution Strategies for Deepfake Detection
• SLA²P: Self-supervised Anomaly Detection with Adversarial Perturbation
• Kernel PCA for Out-of-Distribution Detection: Non-Linear Kernel Selections and Approximations
• Anomaly Detection Based on Critical Paths for Deep Neural Networks
• Training-Free Out-of-Distribution Detection for Pathology Whole-Slide Images
• EarthSynth: Generating Informative Earth Observation with Diffusion Models
• Open Set Label Shift with Test Time Out-of-Distribution Reference
• Spotting the Unexpected (STU): A 3D LiDAR Dataset for Anomaly Segmentation in Autonomous Driving
• Trust The Typical
• Digital Twin-based Out-of-Distribution Detection in Autonomous Vessels
• Dream-Box: Object-wise Outlier Generation for Out-of-Distribution Detection
• Learning to Cascade: Confidence Calibration for Improving the Accuracy and Computational Cost of Cascade Inference Systems
• Enhancing Out-of-Distribution Detection with Extended Logit Normalization
• Proxy-Anchor and EVT-Driven Continual Learning Method for Generalized Category Discovery

Related

• Enhancing The Reliability of Out-of-distribution Image Detection in Neural Networks
• Raising the Bar on the Evaluation of Out-of-Distribution Detection
• Boosting Out-of-distribution Detection with Typical Features
• ImageNet-OOD: Deciphering Modern Out-of-Distribution Detection Algorithms
• Towards Realistic Out-of-Distribution Detection: A Novel Evaluation Framework for Improving Generalization in OOD Detection