Semantic Segmentation in Multiple Adverse Weather Conditions with Domain Knowledge Retention

TL;DR

This paper introduces a novel semantic segmentation approach that effectively adapts to multiple adverse weather conditions by retaining knowledge and improving robustness through pseudo-label blending and weather replay mechanisms.

Contribution

It presents a new method combining adaptive knowledge acquisition, pseudo-label blending, and weather composition replay to enhance continual domain adaptation for adverse weather conditions.

Findings

Achieves an averaged mIoU of 65.7%, outperforming state-of-the-art methods.

Reduces forgetting to 3.6%, significantly lower than previous approaches.

Demonstrates robustness across multiple weather domains on ACDC datasets.

Abstract

Semantic segmentation's performance is often compromised when applied to unlabeled adverse weather conditions. Unsupervised domain adaptation is a potential approach to enhancing the model's adaptability and robustness to adverse weather. However, existing methods encounter difficulties when sequentially adapting the model to multiple unlabeled adverse weather conditions. They struggle to acquire new knowledge while also retaining previously learned knowledge.To address these problems, we propose a semantic segmentation method for multiple adverse weather conditions that incorporates adaptive knowledge acquisition, pseudolabel blending, and weather composition replay. Our adaptive knowledge acquisition enables the model to avoid learning from extreme images that could potentially cause the model to forget. In our approach of blending pseudo-labels, we not only utilize the current model but also integrate the previously learned model into the ongoing learning process. This collaboration between the current teacher and the previous model enhances the robustness of the pseudo-labels for the current target. Our weather composition replay mechanism allows the model to continuously refine its previously learned weather information while simultaneously learning from the new target domain. Our method consistently outperforms the stateof-the-art methods, and obtains the best performance with averaged mIoU (%) of 65.7 and the lowest forgetting (%) of 3.6 against 60.1 and 11.3, on the ACDC datasets for a four-target continual multi-target domain adaptation.