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

TL;DR

This paper introduces GTPC-SSCD, a semi-supervised change detection method that employs two-level perturbation consistency and a gating mechanism to better utilize unlabeled data, outperforming existing methods on benchmark datasets.

Contribution

It proposes a novel two-level perturbation consistency regularization with a gating mechanism for semi-supervised change detection, improving unlabeled data utilization.

Findings

Outperforms seven state-of-the-art methods on six datasets.

Enhances unlabeled data utilization through two-level perturbation consistency.

Gating mechanism effectively assesses sample complexity for targeted perturbations.

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.