面向域自适应分割的掩码表征建模
Masked Representation Modeling for Domain-Adaptive Segmentation
MRM 在潜空间对高层语义特征做掩码重建，可作为通用辅助任务嵌入 DeepLab、DAFormer 等分割网络，训练时联合优化、推理时零开销，为域自适应分割提供了一种简单有效的自监督信号构造方式。
原文依据 We propose Masked Representation Modeling (MRM), an auxiliary task that performs representation masking and reconstruction directly in the latent space.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论语义分割的无监督域自适应旨在将模型从有标注源域迁移至无标注目标域。尽管对比学习等辅助自监督任务增强了特征判别性，但受限于架构约束与目标不一致，掩码建模仍未得到充分探索。本文提出掩码表征建模（MRM），一种直接在潜空间中执行表征掩码与重建的辅助任务。与以往重建低层信号（如像素或视觉 token）的掩码建模方法不同，MRM 以高层语义特征为目标，使其目标与分割任务对齐，并能无缝集成到 DeepLab 和 DAFormer 等标准架构中。为支持高效重建，我们设计了一个轻量级辅助模块 Rebuilder，该模块与分割网络联合训练，但在推理时被移除，从而引入零测试时开销。大量实验表明，MRM 在不同架构和 UDA 基准上均能持续提升分割性能。当与四个代表性基线集成时，MRM 在 GTA -> Cityscapes 上平均提升 +2.3 mIoU，在 Cityscapes -> Synthia 上平均提升 +2.8 mIoU，确立了其作为一种简单、有效且可泛化的无监督域自适应语义分割策略的地位。
英文原文摘要 1404 字符
Unsupervised domain adaptation (UDA) for semantic segmentation seeks to transfer models from a labeled source domain to an unlabeled target domain. While auxiliary self-supervised tasks such as contrastive learning have enhanced feature discriminability, masked modeling remains underexplored due to architectural constraints and misaligned objectives. We propose Masked Representation Modeling (MRM), an auxiliary task that performs representation masking and reconstruction directly in the latent space. Unlike prior masked modeling methods that reconstruct low-level signals (e.g., pixels or visual tokens), MRM targets high-level semantic features, aligning its objective with segmentation and integrating seamlessly into standard architectures like DeepLab and DAFormer. To support efficient reconstruction, we design a lightweight auxiliary module, Rebuilder, which is jointly trained with the segmentation network but removed during inference, introducing zero test-time overhead. Extensive experiments demonstrate that MRM consistently improves segmentation performance across diverse architectures and UDA benchmarks. When integrated with four representative baselines, MRM achieves an average gain of +2.3 mIoU on GTA -> Cityscapes and +2.8 mIoU on Cityscapes -> Synthia, establishing it as a simple, effective, and generalizable strategy for unsupervised domain-adaptive semantic segmentation.