面向开放词表目标检测增强的参数高效语义增强方法
Parameter-Efficient Semantic Augmentation for Enhancing Open-Vocabulary Object Detection
用图像特征金字塔构建多尺度提示词库，为开放词表检测器注入细粒度局部语义，并用语义感知路由器在推理时动态选择增强策略，避免微调破坏预训练泛化性——这套「冻结主干+提示词增强」思路可直接迁移到任何需要跨域迁移的视觉-语言检测或分割模型。
原文依据 we propose a multi-scale prompt bank that leverages image feature pyramids to capture hierarchical semantics
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论开放词表目标检测（OVOD）使模型能够检测任意物体类别，包括未见过的类别。得益于大规模预训练，现有 OVOD 方法在通用场景（如 OV-COCO）上取得了强劲的检测性能，但在迁移到存在显著域偏移的下游任务时性能会严重下降。这种退化源于特定领域任务中类别标签的稀缺性与弱语义性，以及现有模型无法捕获粗粒度类别标签之外的辅助语义。为解决这些问题，我们提出 HSA-DINO，一个用于增强开放词表目标检测的参数高效语义增强框架。具体而言，我们提出多尺度提示词库，利用图像特征金字塔捕获层次化语义并选取领域特定的局部语义提示词，从粗到细地逐步丰富文本表示。此外，我们引入语义感知路由器，在推理时动态选择适当的语义增强策略，从而防止参数更新损害预训练 OVOD 模型的泛化能力。我们在 OV-COCO、若干垂直领域数据集以及修改后的基准设置上评估了 HSA-DINO。结果表明，HSA-DINO 优于此前的最优方法，在领域适应性与开放词表泛化之间取得了更优的权衡。
英文原文摘要 1512 字符
Open-vocabulary object detection (OVOD) enables models to detect any object category, including unseen ones. Benefiting from large-scale pre-training, existing OVOD methods achieve strong detection performance on general scenarios (e.g., OV-COCO) but suffer severe performance drops when transferred to downstream tasks with substantial domain shifts. This degradation stems from the scarcity and weak semantics of category labels in domain-specific task, as well as the inability of existing models to capture auxiliary semantics beyond coarse-grained category label. To address these issues, we propose HSA-DINO, a parameter-efficient semantic augmentation framework for enhancing open-vocabulary object detection. Specifically, we propose a multi-scale prompt bank that leverages image feature pyramids to capture hierarchical semantics and select domain-specific local semantic prompts, progressively enriching textual representations from coarse to fine-grained levels. Furthermore, we introduce a semantic-aware router that dynamically selects the appropriate semantic augmentation strategy during inference, thereby preventing parameter updates from degrading the generalization ability of the pre-trained OVOD model. We evaluate HSA-DINO on OV-COCO, several vertical domain datasets, and modified benchmark settings. The results show that HSA-DINO performs favorably against previous state-of-the-art methods, achieving a superior trade-off between domain adaptability and open-vocabulary generalization.