ReSAM：精炼、重查询与强化——面向遥感图像的自提示词点监督分割
ReSAM: Refine, Requery, and Reinforce: Self-Prompting Point-Supervised Segmentation for Remote Sensing Images
把 SAM 的交互式提示词从人工输入改为自生成循环（点→伪掩码→自构框提示词→重查询），并用软语义对齐约束嵌入，这套自提示词自训练流程可直接迁移到任何缺乏密集标注的专业成像域分割任务。
原文依据 coarse pseudo-masks are generated from initial points (Refine), improved with self-constructed box prompts (Requery)
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论诸如 Segment Anything Model（SAM）之类的交互式分割模型在自然图像上展现出显著的泛化能力，但由于严重的域偏移与密集标注的稀缺，它们在遥感图像（RSI）上表现欠佳。为克服这一局限，我们提出一种点监督的自提示词框架，仅使用稀疏点标注即可将 SAM 适配到遥感图像。该方法采用「精炼—重查询—强化」循环：由初始点生成粗糙伪掩码（精炼），借助自构造的框提示词对其改进（重查询），并通过软语义对齐（SSA）对齐嵌入以缓解误差传播（强化）。在不依赖全掩码监督的情况下，我们的方法通过自引导的提示词适配逐步提升 SAM 的分割质量与域鲁棒性。我们在 WHU、HRSID 与 NWPU VHR-10 三个遥感基准数据集上评估所提方法，结果表明其持续优于预训练 SAM 以及近期的点监督分割方法。与全监督模型相比，在仅使用单点标注的情况下，我们的方法将性能差距缩小至 1.3%（WHU）、4.9%（HRSID）与 8.5%（NWPU）。结果说明，自提示词与语义对齐为面向遥感应用的基础分割模型提供了一条高效的可扩展点级适配路径。
英文原文摘要 1435 字符
Interactive segmentation models such as the Segment Anything Model (SAM) have demonstrated remarkable generalization on natural images, but they perform suboptimally on remote sensing imagery (RSI) due to severe domain shifts and the scarcity of dense annotations. To address this limitation, we propose a point-supervised, self-prompting framework that adapts SAM to RSI using only sparse point annotations. Our method employs a Refine-Requery-Reinforce loop, in which coarse pseudo-masks are generated from initial points (Refine), improved with self-constructed box prompts (Requery), and embeddings are aligned with Soft Semantic Alignment (SSA) to mitigate error propagation. (Reinforce). Without relying on full-mask supervision, our approach progressively enhances SAM's segmentation quality and domain robustness through self-guided prompt adaptation. We evaluate our proposed method on three RSI benchmark datasets, WHU, HRSID, and NWPU VHR-10, demonstrating that it consistently outperforms pretrained SAM and recent point-supervised segmentation methods. Compared to the fully supervised model, our approach reduces the performance gap to 1.3% (WHU), 4.9% (HRSID), and 8.5% (NWPU) while relying only on 1-point annotations. Our results demonstrate that self-prompting and semantic alignment provide an efficient path towards scalable, point-level adaptation of foundation segmentation models for remote sensing applications.