Perceive, Act and Correct: Confidence Is Not Enough for Hyperspectral Classification

2025-11-13

AI Summary

In hyperspectral image classification, models relying solely on high-confidence predictions suffer from confirmation bias—particularly under sparse labeling and class imbalance—leading to overfitting on erroneous predictions and degraded generalization. To address this, we propose CABIN, a novel framework establishing a “perception–decision–correction” closed-loop learning paradigm. CABIN introduces an uncertainty-guided dual-sampling strategy and a fine-grained dynamic allocation mechanism to precisely identify and rectify ambiguous or noisy samples. It further integrates a cognition-inspired semi-supervised learning scheme, combining cognitive uncertainty estimation, dynamically refined pseudo-labeling, and a customized loss function. Extensive experiments demonstrate that CABIN consistently boosts classification accuracy and annotation efficiency across multiple state-of-the-art methods. Notably, it exhibits superior robustness under low-labeling-ratio and long-tailed distribution scenarios, achieving significant improvements in both predictive reliability and label efficiency.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Representation Learning for Vision
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Confidence alone is often misleading in hyperspectral image classification, as models tend to mistake high predictive scores for correctness while lacking awareness of uncertainty. This leads to confirmation bias, especially under sparse annotations or class imbalance, where models overfit confident errors and fail to generalize. We propose CABIN (Cognitive-Aware Behavior-Informed learNing), a semi-supervised framework that addresses this limitation through a closed-loop learning process of perception, action, and correction. CABIN first develops perceptual awareness by estimating epistemic uncertainty, identifying ambiguous regions where errors are likely to occur. It then acts by adopting an Uncertainty-Guided Dual Sampling Strategy, selecting uncertain samples for exploration while anchoring confident ones as stable pseudo-labels to reduce bias. To correct noisy supervision, CABIN introduces a Fine-Grained Dynamic Assignment Strategy that categorizes pseudo-labeled data into reliable, ambiguous, and noisy subsets, applying tailored losses to enhance generalization. Experimental results show that a wide range of state-of-the-art methods benefit from the integration of CABIN, with improved labeling efficiency and performance.

Problem

• Addresses misleading confidence in hyperspectral image classification models
• Solves confirmation bias from sparse annotations and class imbalance
• Corrects noisy supervision through uncertainty-guided learning strategies

Innovation

• Semi-supervised framework with closed-loop learning process
• Uncertainty-guided dual sampling strategy for exploration
• Fine-grained dynamic assignment strategy for noisy supervision