Lightweight Adapter Learning for More Generalized Remote Sensing Change Detection

TL;DR

This work tackles the generalization gap in remote sensing change detection (CD) by introducing CANet, a two-component network with a dataset-shared backbone and a lightweight dataset-specific adapter. A key innovation is the ICM change region mask, which adaptively focuses on relevant change objects, alongside per-dataset batch normalization to mitigate distribution differences. CANet achieves competitive cross-dataset CD performance while requiring updates to only a small fraction (about 4.1%–7.7%) of parameters when adapting to new datasets, and demonstrates robustness under limited training data and with online adaptation. The approach provides strong practical value for deploying CD models across diverse datasets with reduced training costs and improved generalization.

Abstract

Deep learning methods have shown promising performances in remote sensing image change detection (CD). However, existing methods usually train a dataset-specific deep network for each dataset. Due to the significant differences in the data distribution and labeling between various datasets, the trained dataset-specific deep network has poor generalization performances on other datasets. To solve this problem, this paper proposes a change adapter network (CANet) for a more universal and generalized CD. CANet contains dataset-shared and dataset-specific learning modules. The former explores the discriminative features of images, and the latter designs a lightweight adapter model, to deal with the characteristics of different datasets in data distribution and labeling. The lightweight adapter can quickly generalize the deep network for new CD tasks with a small computation cost. Specifically, this paper proposes an interesting change region mask (ICM) in the adapter, which can adaptively focus on interested change objects and decrease the influence of labeling differences in various datasets. Moreover, CANet adopts a unique batch normalization layer for each dataset to deal with data distribution differences. Compared with existing deep learning methods, CANet can achieve satisfactory CD performances on various datasets simultaneously. Experimental results on several public datasets have verified the effectiveness and advantages of the proposed CANet on CD. CANet has a stronger generalization ability, smaller training costs (merely updating 4.1%-7.7% parameters), and better performances under limited training datasets than other deep learning methods, which also can be flexibly inserted with existing deep models.

Figures (4)

• Figure 1: Existing deep learning methods usually train a dataset-specific deep network for different datasets, but their generalization performance is poor on other datasets. This paper proposes a change adapter network for more generalized change detection, which can perform well on various datasets simultaneously.
• Figure 2: Framework of the proposed change adapter network, CANet, for change detection. CANet contains a dataset-shared learning module and a dataset-specific learning module, adapter.
• Figure 3: F1 results of SYSU, LEVIR, and WHU datasets based on different numbers of training samples, i.e., 10% and 100%.
• Figure 4: The results of the original deep models on the historical dataset (CDD) and the results of the updated deep models through the online training strategy on the historical (CDD+FT) and new datasets (SYSU+FT, LEVIR+FT, and WHU+FT). FT means the deep model is trained on the CDD dataset and then fine-tuned on the new dataset.