HySeg：为结构感知遥感分割学习生成式先验
HySeg: Learning Generative Priors for Structure-Aware Remote Sensing Segmentation
把分割后验推理建立在生成式结构先验之上：用 MeanFlow 学习语义拓扑的连续随机场，再投影成类别条件亲和度去引导分割头，这个「先验→亲和度→后验」的插件式设计可以迁移到任何需要边界连贯性的密集预测任务。
原文依据 reformulating RSISS as posterior inference grounded in generative structural priors
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论高分辨率遥感影像呈现出复杂的空间规律性，其中拓扑、连续性与区域邻接关系支配着语义组织。然而，现有的遥感图像语义分割（RSISS）网络以判别式为主，从数据中估计出强后验，却缺乏能够编码此类结构依赖关系的生成式先验。这种失衡导致边界破碎、纹理过拟合以及跨数据集泛化能力差。为应对这一挑战，我们将 RSISS 重新表述为以生成式结构先验为基础的后验推理，提出 HySeg——一种混合生成-判别式分割范式，通过生成建模学习结构一致的先验，并引导遥感分割的后验推理。其核心是 MeanStruct 模块，这是一个基于 MeanFlow 的生成式先验学习器，将语义拓扑建模为连续随机场；而先验到亲和度投影（P2A）则动态地将该场转换为拓扑感知的、类别条件的亲和度，在动态亲和度驱动分割（DAS）头中引导后验推理。我们的方法是模型无关的，可与多种骨干网络无缝集成，并持续提升结构连贯性与泛化能力。在四个具有挑战性的 RSISS 基准上，HySeg 取得了最优性能，并推动遥感分割从基于外观的感知走向结构推理。我们的代码已发布于 https://github.com/HeryJie/HySeg。
英文原文摘要 1554 字符
High-resolution remote sensing imagery exhibits complex spatial regularities where topology, continuity, and region adjacency govern semantic organization. However, existing remote sensing image semantic segmentation (RSISS) networks, being predominantly discriminative, estimate strong posteriors from data while lacking generative priors that encode such structural dependencies. This imbalance leads to fragmented boundaries, texture overfitting, and poor cross-dataset generalization. We address this challenge by reformulating RSISS as posterior inference grounded in generative structural priors, introducing HySeg , a hybrid generative-discriminative segmentation paradigm that learns structure-consistent priors through generative modeling and guides posterior inference for remote sensing segmentation. At its core, the MeanStruct module, a MeanFlow-based generative prior learner, models semantic topology as a continuous stochastic field, while the Prior-to-Affinity Projection (P2A) dynamically transforms this field into topology-aware, class-conditional affinities that guide posterior inference in the Dynamic Affinity-driven Segmentation (DAS) head. Our approach is model-agnostic and seamlessly integrates with diverse backbones, consistently improving structural coherence and generalization. Across four challenging RSISS benchmarks, HySeg achieves state-of-the-art performance and advances remote sensing segmentation from appearance-based perception to structural reasoning. Our code is available at https://github.com/HeryJie/HySeg.