ICM：恶劣天气下域适应的类内混合
ICM: Intra-class Mixing for Domain Adaptation in Adverse Weather
浏览论文内容
中文总结 AI 辅助
提出类内混合一致性（ICM）框架，通过在同一图像和语义类内混合增强预测一致性，解决恶劣天气下无监督域适应中伪标签不可靠的问题，在Cityscapes到ACDC基准上达到75.7% mIoU，刷新最先进性能。
中文摘要 AI 辅助
在恶劣天气条件下，无监督域适应（UDA）用于语义分割仍然具有挑战性，因为严重的外观变化扩大了域差距，并降低了目标域中伪标签的可靠性。为了解决这个问题，我们提出了一个类内混合一致性（ICM）框架，该框架强制类内混合图像与其原始对应图像之间的预测一致性。与以往基于混合的一致性方法不同，这些方法结合了不同图像或域的区域，可能引入不切实际的语义不一致性，ICM在同一图像和语义类内进行混合，保留了用于一致性正则化的真实语义布局。通过ICM，我们在语义分割的晴朗到恶劣天气无监督域适应（UDA）中建立了新的最先进性能。在Cityscapes $\ ightarrow$ ACDC基准测试中，我们的方法达到了75.7%的mIoU，比之前的最先进水平高出+1.9个百分点，证明了其在缓解具有挑战性的环境条件下的类混淆方面的有效性。代码在补充材料中提供。
英文摘要
Unsupervised domain adaptation (UDA) for semantic segmentation remains challenging under adverse weather conditions because severe appearance changes enlarge the domain gap and degrade the reliability of pseudo labels in the target domain. To address this problem, we propose an Intra-Class Mixing Consistency (ICM) framework that enforces prediction consistency between an intra-class mixed image and its original counterpart. Unlike previous mixing-based consistency methods that combine regions across different images or domains and may introduce unrealistic semantic inconsistencies, ICM performs mixing within the same image and semantic class, preserving realistic semantic layout for consistency regularization. With ICM, we establish a new state-of-the-art performance for clear-to-adverse-weather unsupervised domain adaptation (UDA) in semantic segmentation. On the Cityscapes $\rightarrow$ ACDC benchmark, our method achieves 75.7\% mIoU, outperforming the previous state of the art by +1.9 pp, demonstrating its effectiveness in mitigating class confusion under challenging environmental conditions. The code is provided in the supplementary material.
