Source-Free Online Domain Adaptive Semantic Segmentation of Satellite Images under Image Degradation
This work addresses the problem of adapting semantic segmentation models to degraded satellite images in real-time for remote sensing applications, representing an incremental improvement in domain adaptation techniques.
The paper tackled online domain adaptation for satellite image segmentation under image degradation by proposing a test-time adaptation method that progressively estimates batch normalization statistics and refines masks using class centers, achieving robust performance across various scenarios.
Online adaptation to distribution shifts in satellite image segmentation stands as a crucial yet underexplored problem. In this paper, we address source-free and online domain adaptation, i.e., test-time adaptation (TTA), for satellite images, with the focus on mitigating distribution shifts caused by various forms of image degradation. Towards achieving this goal, we propose a novel TTA approach involving two effective strategies. First, we progressively estimate the global Batch Normalization (BN) statistics of the target distribution with incoming data stream. Leveraging these statistics during inference has the ability to effectively reduce domain gap. Furthermore, we enhance prediction quality by refining the predicted masks using global class centers. Both strategies employ dynamic momentum for fast and stable convergence. Notably, our method is backpropagation-free and hence fast and lightweight, making it highly suitable for on-the-fly adaptation to new domain. Through comprehensive experiments across various domain adaptation scenarios, we demonstrate the robust performance of our method.