Locally Consistent Transductive Information Maximization for Few-Shot Remote Sensing Scene Classification

2026-07-31

AI Summary

This work addresses the underutilization of unlabeled query set structure in few-shot remote sensing image classification by extending TIM++ with a local consistency regularizer that enforces prediction agreement between each query sample and its κ nearest neighbors in feature space. The proposed method integrates a CLIP-driven vision-language model, transductive information maximization, neighbor graph construction, and a multi-source affinity graph fusion strategy. Evaluated on ten remote sensing datasets, it significantly outperforms existing approaches, demonstrating particularly strong performance under extremely low-shot settings. Additionally, the authors release the first open-source benchmark for transductive few-shot classification in remote sensing.

Abstract

Remote sensing scene classification is increasingly relying on foundation models pre-trained on large-scale Earth-observation data. Moreover, transductive inference, which exploits the collective statistical structure of the entire unlabeled query set, appears to naturally match remote sensing pipelines where large images are routinely split into patches and inferred as a batch. In this work, we introduce LC-TIM (Locally Consistent Transductive Information Maximization), which extends the state-of-the-art Transductive Information Maximization for Few-Shot CLIP (TIM++) objective with a local consistency regularizer that enforces prediction agreement between each query sample and its $κ$ nearest feature-space neighbors. The regularizer enters as a single multiplicative factor in the closed-form $q$-update, adding negligible computational overhead. We further propose a multi-source extension that fuses the affinity graph from multiple remote sensing foundation model, further boosting classification accuracy. To assess these methods, we establish the first comprehensive, open-source benchmark for transductive few-shot RS scene classification, evaluating LP++, TransCLIP, TIM++, and LC-TIM across ten diverse datasets, two remote sensing vision-language models, and across various few-shot settings. Our experiments show that transductive methods consistently outperform zero-shot baselines, and that LC-TIM achieves state-of-the-art accuracy, with the largest gains in the low-shot regime where neighborhood cues are most informative. Code is publicly available at: https://github.com/elkhouryk/LC-TIM

Problem

• few-shot learning
• remote sensing scene classification
• transductive inference
• foundation models
• local consistency

Innovation

• Transductive Learning
• Few-Shot Classification
• Local Consistency Regularization
• Remote Sensing Scene Classification
• Foundation Models