GTPC-SSCD: Gate-guided Two-level Perturbation Consistency-based Semi-Supervised Change Detection

2024-11-28

arXiv.org

AI Summary

To address insufficient utilization of unlabeled remote sensing images in semi-supervised change detection (SSCD), this paper proposes a joint constraint framework integrating image-level strong-weak consistency and feature-level dual-layer perturbation consistency. We innovatively design a gated guidance mechanism for dual-layer perturbations, which dynamically determines whether to apply feature-level perturbations based on sample difficulty, thereby enabling differentiated and efficient exploitation of unlabeled data. The method jointly incorporates multi-scale feature perturbations and gated hardness estimation to enhance model robustness against pseudo-label noise. Evaluated on six benchmark remote sensing datasets, our approach achieves an average 3.2% improvement in F1-score over seven state-of-the-art methods, demonstrating the effectiveness and generalizability of both the dual-layer consistency modeling and the adaptive gating mechanism.

Technology Category

• Machine Learning: Semi-Supervised Learning
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Semi-supervised change detection (SSCD) utilizes partially labeled data and abundant unlabeled data to detect differences between multi-temporal remote sensing images. The mainstream SSCD methods based on consistency regularization have limitations. They perform perturbations mainly at a single level, restricting the utilization of unlabeled data and failing to fully tap its potential. In this paper, we introduce a novel Gate-guided Two-level Perturbation Consistency regularization-based SSCD method (GTPC-SSCD). It simultaneously maintains strong-to-weak consistency at the image level and perturbation consistency at the feature level, enhancing the utilization efficiency of unlabeled data. Moreover, we develop a hardness analysis-based gating mechanism to assess the training complexity of different samples and determine the necessity of performing feature perturbations for each sample. Through this differential treatment, the network can explore the potential of unlabeled data more efficiently. Extensive experiments conducted on six benchmark CD datasets demonstrate the superiority of our GTPC-SSCD over seven state-of-the-art methods.

Problem

• Improves semi-supervised change detection using unlabeled data
• Enhances perturbation consistency at image and feature levels
• Introduces gating mechanism for efficient sample complexity analysis

Innovation

• Two-level perturbation consistency for SSCD
• Gate-guided hardness analysis mechanism
• Enhanced unlabeled data utilization efficiency