Skip to content
Preprint

See the Future, Classify the Present: Stochastic Latent World Modeling for Hyperspectral Images

Oct 2026 · 0 citations · 61 references
Computer Science

TL;DR

Stochastic Latent World Modeling (LWM), a trajectory-based representation-learning framework that enriches classification through stochastic latent-state prediction, enables the classifier to exploit structured dependencies in the representation space and provides richer features for HSIC.

Abstract

Hyperspectral image classification (HSIC) relies on static spatial-spectral representations learned from labeled samples, limiting its ability to model structured dependencies in the latent representation space. To address this limitation, this work proposes Stochastic Latent World Modeling (LWM), a trajectory-based representation-learning framework that enriches classification through stochastic latent-state prediction. Given a hyperspectral patch, LWM tokenizes and encodes spectral and spatial information separately using dedicated transformer branches, then fuses their outputs into a unified representation. This representation is further organized into an environment state that captures center, neighborhood, spectral, and global contextual information. A stochastic latent transition module then predicts multiple future latent states through Gaussian parameterization and reparameterization, with recurrent refinement used to model dependencies across successive states. The resulting latent trajectory is adaptively aggregated by a rollout planner to construct a planned representation, which is integrated with the base representation for final classification. By extending static spatial-spectral encoding with stochastic multi-step latent-state modeling, LWM enables the classifier to exploit structured dependencies in the representation space and provides richer features for HSIC. Across five benchmark datasets, LWM achieves overall accuracies ranging from 89.21% to 99.52%, demonstrating consistent classification performance under limited-label settings.

View source

Similar papers

Open access Oct 2026

Dual-Domain Triple-View Contrastive Representation Learning for Few-Shot Hyperspectral Image Classification

Few-shot hyperspectral image classification (FS-HSIC) remains highly challenging because the limited number of labeled samples not only makes traditional supervised learning models prone to overfitting but also restricts their ability to learn generalizable spatial–spectral representations. Moreover, existing methods a...

Ru-Jun Zhang, Wen-Ming Cao, Qi-Fan Liu · 0 citations
Open access Oct 2026

Cross-Task Latent Feature Distillation Learning for Hyperspectral Anomaly Detection

A novel knowledge transfer paradigm from classification tasks to anomaly detection is proposed, which employs a pretrained convolutional enhanced Transformer (CET) classification network as the teacher model, transferring its rich spectral–spatial context encodings via feature-level distillation.

Lei Xia, Yin-Nian Liu · 0 citations
Open access Oct 2026

Spatial contextual learning for hyperspectral imagery classification using transformer encoder framework based ensemble model

This work proposes a four–stage hybrid framework for hyperspectral image classification that integrates spectral modeling and spatial feature extraction within a unified pipeline and compares the model with recent baselines, including DaViT.

Disha Goyal, Rashi Mishra, Mohit Somani et al. · 0 citations
Open access 2026

HCD2: A Generative Dense Classification Framework for Hyperspectral Imagery With Conditional Diffusion

A hyperspectral conditional diffusion for dense (HCD2) classification framework is proposed in this article, which reformulates HSI classification as a conditional label-map denoising process guided by hyperspectral features, enabling generative dense semantic inference.

Hong-Qiao Wang, Hai-Yan Sun, Xuan-Chen Yan et al. · 0 citations
Conference Oct 2026

Remote sensing hierarchical conditional transformer for hyperspectral image classification

RS-HCFormer is proposed, a hierarchical conditional encoding Transformer for hyperspectral image classification that substantially outperforms state-of-the-art methods in both accuracy and efficiency.

Qiang Li, Hao-Yue Xu, Jia-Ning Zhang et al. · 0 citations
Open access 2026

UST-Net: Unified Spatial–Spectral–Temporal Modeling for Adaptive Hyperspectral Object Tracking

Hyperspectral (HS) imagery provides material-sensitive spectral cues that complement spatial appearance and improve target discrimination when red–green–blue (RGB) observations are ambiguous. Existing HS trackers have explored spectral prompting and temporal modeling, yet spectrum-preserving prompt construction, recurr...

Yan-Ni Dong, Jia-Qi Zhang, Xiao-Le Han et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.