HOpenCls: Training Hyperspectral Image Open-Set Classifiers in Their Living Environments

2025-02-21

AI Summary

Open-set classification of hyperspectral images (HSI) faces two key challenges: reliance on manually annotated unknown samples and poor adaptability to real-world, mixed野外 data with unknown class distributions. Method: This paper pioneers formulating HSI open-set classification as a positive-unlabeled (PU) learning problem. We propose a multi-label strategy to jointly optimize known-class recognition and unknown-class rejection; design gradient contraction and expansion modules to mitigate optimization bias caused by anomalous gradients in野外 data; and integrate hyperspectral feature embedding with open-set decision boundary modeling. Contribution/Results: Evaluated on multiple HSI benchmarks, our method achieves absolute improvements of 12.3% in F1-score and 9.8% in open-set classification rate (OSCR), significantly enhancing robust unknown-class rejection under unlabeled and heterogeneous scenarios.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Learning to Search

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Hyperspectral image (HSI) open-set classification is critical for HSI classification models deployed in real-world environments, where classifiers must simultaneously classify known classes and reject unknown classes. Recent methods utilize auxiliary unknown classes data to improve classification performance. However, the auxiliary unknown classes data is strongly assumed to be completely separable from known classes and requires labor-intensive annotation. To address this limitation, this paper proposes a novel framework, HOpenCls, to leverage the unlabeled wild data-that is the mixture of known and unknown classes. Such wild data is abundant and can be collected freely during deploying classifiers in their living environments. The key insight is reformulating the open-set HSI classification with unlabeled wild data as a positive-unlabeled (PU) learning problem. Specifically, the multi-label strategy is introduced to bridge the PU learning and open-set HSI classification, and then the proposed gradient contraction and gradient expansion module to make this PU learning problem tractable from the observation of abnormal gradient weights associated with wild data. Extensive experiment results demonstrate that incorporating wild data has the potential to significantly enhance open-set HSI classification in complex real-world scenarios.

Problem

• Open-set classification for hyperspectral images
• Utilizing unlabeled wild data for training
• Reformulating classification as a PU learning problem

Innovation

• Utilizes unlabeled wild data
• Reformulates as PU learning
• Introduces gradient contraction-expansion