Generalized Category Discovery in Hyperspectral Images via Prototype Subspace Modeling

2025-09-28

AI Summary

This work addresses the generalized category discovery (GCD) problem for hyperspectral imagery (HSI), where conventional RGB-oriented GCD methods fail due to HSI’s high dimensionality, strong spectral correlations, and complex spectral structures. We present the first GCD framework tailored to HSI, introducing a prototype subspace modeling approach: instead of representing each class with a single point prototype, we model both known and novel categories using learnable, orthogonal basis vectors that span a low-dimensional subspace. To ensure geometric interpretability and spectral fidelity, we impose orthogonality constraints on the basis and reconstruction constraints on sample projections—optimized via self-supervised learning. Evaluated on real-world hyperspectral datasets, our method significantly outperforms existing GCD approaches. Ablations confirm its enhanced capacity for discriminative spectral representation learning and semantic category disentanglement. This work establishes a new paradigm for unsupervised semantic discovery in HSI, bridging a critical gap between spectral data analysis and open-world visual recognition.

Technology Category

• Machine Learning: Semi-Supervised Learning
• Computer Vision: Other Foundations of Computer Vision
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

Generalized category discovery~(GCD) seeks to jointly identify both known and novel categories in unlabeled data. While prior works have mainly focused on RGB images, their assumptions and modeling strategies do not generalize well to hyperspectral images~(HSI), which are inherently high-dimensional and exhibit complex spectral structures. In this paper, we propose the first GCD framework tailored for HSI, introducing a prototype subspace modeling model to better capture class structure. Instead of learning a single prototype vector for each category as in existing methods such as SimGCD, we model each category using a set of basis vectors, forming a subspace representation that enables greater expressiveness and discrimination in a high-dimensional feature space. To guide the learning of such bases, we enforce two key constraints: (1) a basis orthogonality constraint that promotes inter-class separability, and (2) a reconstruction constraint that ensures each prototype basis can effectively reconstruct its corresponding class samples. Experimental results on real-world HSI demonstrate that our method significantly outperforms state-of-the-art GCD methods, establishing a strong foundation for generalized category discovery in hyperspectral settings.

Problem

• Identifying known and novel categories in unlabeled hyperspectral image data
• Modeling categories via subspace prototypes for high-dimensional spectral discrimination
• Enhancing class separability and reconstruction in hyperspectral category discovery

Innovation

• Subspace modeling replaces single prototype vectors
• Orthogonality constraint enhances inter-class separability
• Reconstruction constraint ensures effective sample representation