BiPA：面向水下实例分割的双层提示词适配
BiPA: Bilevel Prompt Adaptation for Underwater Instance Segmentation
把 SAM 的稠密提示词学习建模成双层优化，并用贝叶斯优化在冻结主干的前提下先搜提示词、再微调模型，这套「先优化提示词、后调参数」的两阶段流程可直接迁移到其他严重域偏移场景（如医学、遥感）的 SAM 适配。
原文依据 We formulate dense prompt learning as a bilevel optimization, explicitly capturing the mutual dependency between prompt and model.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论水下实例分割对于细粒度场景理解至关重要。然而，由于严重的退化（如水体浑浊），水下图像与空中视觉之间存在显著的域差异。因此，尽管 SAM 具备通用分割能力，其在水下场景中的性能却急剧下降。本文提出 BiPA，可有效地将 SAM 适配到水下域。具体而言，我们构建了带有双提示词的水下 SAM，并引入前景注意力注入模块以增强局部前景表征。我们将稠密提示词学习建模为双层优化，显式刻画提示词与模型之间的相互依赖关系。为使该问题可求解，我们设计了两阶段学习策略。第一阶段对稠密提示词本身进行适配，利用贝叶斯优化对其进行更新以实现高效学习。第二阶段在冻结已优化提示词的条件下微调模型参数，最终实现有效的跨域适配。大量实验与分析验证了 BiPA 的优越性与高效性。代码已公开于 https://github.com/ZeAstra/BiPA。
英文原文摘要 1136 字符
Underwater instance segmentation is essential for fine-grained scene understanding. However, underwater imagery exhibits a strong domain gap from in-air vision due to severe degradation (e.g., turbidity). Consequently, despite its general segmentation ability, SAM degrades sharply underwater. In this work, we propose BiPA, which effectively adapts SAM to the underwater domain. To be concrete, we construct an underwater SAM with dual prompts and introduce a foreground-attentive injection block to enhance local foreground representation. We formulate dense prompt learning as a bilevel optimization, explicitly capturing the mutual dependency between prompt and model. To make this tractable, we design a two-stage learning strategy. The first stage adapts the dense prompt itself, updating it with Bayesian optimization to learn efficiently. The second stage fine-tunes the model parameters under the frozen optimized prompt, which finally enables effective cross-domain adaptation. Extensive experiments and analyses verify the superiority and efficiency of BiPA. The code is publicly available at https://github.com/ZeAstra/BiPA.