Known Meets Unknown: Mitigating Overconfidence in Open Set Recognition

2025-11-15

AI Summary

In open-set recognition (OSR), models often exhibit overconfidence on semantically proximal unknown samples, misclassifying them as known classes and degrading rejection capability. To address this, we propose a two-stage decoupled framework: Stage I quantifies sample-level uncertainty via controlled feature perturbations that generate diverse predictions; Stage II employs a dual-classifier architecture jointly optimizing known-class classification and unknown-class detection, explicitly suppressing overconfidence induced by inter-class feature overlap. Our method requires no prior knowledge of unknown classes and is end-to-end trainable. Evaluated on CIFAR-10, CIFAR-100, and TinyImageNet benchmarks, it significantly outperforms state-of-the-art OSR approaches, achieving superior performance in key metrics—including FPR95 and AUROC—while enhancing the separability and robustness of decision boundaries.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Uncertainty Representations
• Computer Vision: Object Detection & Categorization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Open Set Recognition (OSR) requires models not only to accurately classify known classes but also to effectively reject unknown samples. However, when unknown samples are semantically similar to known classes, inter-class overlap in the feature space often causes models to assign unjustifiably high confidence to them, leading to misclassification as known classes -- a phenomenon known as overconfidence. This overconfidence undermines OSR by blurring the decision boundary between known and unknown classes. To address this issue, we propose a framework that explicitly mitigates overconfidence caused by inter-class overlap. The framework consists of two components: a perturbation-based uncertainty estimation module, which applies controllable parameter perturbations to generate diverse predictions and quantify predictive uncertainty, and an unknown detection module with distinct learning-based classifiers, implemented as a two-stage procedure, which leverages the estimated uncertainty to improve discrimination between known and unknown classes, thereby enhancing OSR performance. Experimental results on three public datasets show that the proposed framework achieves superior performance over existing OSR methods.

Problem

• Mitigating overconfidence in open set recognition for unknown samples
• Reducing misclassification when unknown classes resemble known ones
• Improving discrimination between known and unknown classes using uncertainty estimation

Innovation

• Perturbation-based uncertainty estimation for predictive confidence
• Two-stage unknown detection with learning-based classifiers
• Mitigating overconfidence by leveraging inter-class overlap uncertainty