Reshaping the Online Data Buffering and Organizing Mechanism for Continual Test-Time Adaptation
This work addresses the challenges of error accumulation and catastrophic forgetting in CTTA, which is crucial for deploying AI models in dynamic real-world environments, though it appears to be an incremental improvement over existing methods.
The paper tackles the problem of adapting pre-trained models to continually changing unsupervised target domains in Continual Test-Time Adaptation (CTTA) by reshaping online data buffering and organizing mechanisms, resulting in superior performance in segmentation and classification tasks as demonstrated through extensive experiments.
Continual Test-Time Adaptation (CTTA) involves adapting a pre-trained source model to continually changing unsupervised target domains. In this paper, we systematically analyze the challenges of this task: online environment, unsupervised nature, and the risks of error accumulation and catastrophic forgetting under continual domain shifts. To address these challenges, we reshape the online data buffering and organizing mechanism for CTTA. We propose an uncertainty-aware buffering approach to identify and aggregate significant samples with high certainty from the unsupervised, single-pass data stream. Based on this, we propose a graph-based class relation preservation constraint to overcome catastrophic forgetting. Furthermore, a pseudo-target replay objective is used to mitigate error accumulation. Extensive experiments demonstrate the superiority of our method in both segmentation and classification CTTA tasks. Code is available at https://github.com/z1358/OBAO.