C2F2: Cross-Task Cross-Domain Feature Fusion for Semi-Supervised Change Detection

Dongjie Zhang,Yuting Hong,Xiaojie Qiu,Li Dong,Diqun Yan,Chengbin Peng

DOI: https://doi.org/10.1109/lgrs.2024.3467260

2024-01-01

Abstract:Semi-supervised learning for change detection (CD), which significantly reduces the labor costs associated with data annotation, has recently garnered substantial attention. In this study, we propose to enhance traditional semi-supervised learning frameworks by leveraging cross-task cross-domain (CTCD) models, which generate complementary features that differ from standard hidden features. The procedure is as follows. First, the standard features obtained from a traditional encoding-decoding structure are fused with attention-augmented complementary features. Second, a secondary decoder maps the fused heterogeneous features into the label space to obtain high-quality pseudo-labels, offering more precise guidance for semi-supervised learning on traditional structures. This approach improves pseudo-labels by leveraging the strength of CTCD models, including large pretrained models, to enhance the semi-supervised learning process of domain-specific and task-specific models. Experimental results on benchmark datasets demonstrate that our proposed approach surpasses state-of-the-art methods.