DRA-SAM: Disagreement-Driven Referential Active SAM for Low-Altitude Drone Referring Segmentation

Computer Science > Computer Vision and Pattern Recognition arXiv:2610.33057 (cs) [Submitted on 19 Oct 2026] Title: DRA-SAM: Disagreement-Driven Referential Active SAM for Low-Altitude Drone Referring Segmentation Authors: Jisheng Dang, Huaiyuan Peng, Wencan Zhang, Bimei Wang, Jingwen Zhao, Wenjun Ma, Hong Peng, Bin Hu, Qi Tian, Tat-Seng Chua.

计算机科学 > 计算机视觉与模式识别 arXiv:2610.33057 (cs) [提交于 2026 年 10 月 19 日] 标题：DRA-SAM：用于低空无人机指代分割的差异驱动主动式 SAM。作者：Jisheng Dang, Huaiyuan Peng, Wencan Zhang, Bimei Wang, Jingwen Zhao, Wenjun Ma, Hong Peng, Bin Hu, Qi Tian, Tat-Seng Chua。

Abstract: Referring segmentation in low-altitude drone imagery requires distinguishing small, visually similar instances under complex linguistic and spatial constraints. Global inference may yield multiple plausible masks or leave the intended referent insufficiently represented, motivating selective local re-inference.

摘要：低空无人机图像中的指代分割需要在复杂的语言和空间约束下区分细小且视觉相似的实例。全局推理可能会产生多个看似合理的掩码，或者导致目标指代对象未能得到充分表达，这促使了选择性局部重推理的需求。

We propose DRA-SAM, a disagreement-driven framework that allocates an additional inference pass according to competing referential hypotheses and an unresolved state. A lightweight EvidenceScorer estimates beliefs over existing candidates and the unresolved state.

我们提出了 DRA-SAM，这是一个差异驱动的框架，它根据相互竞争的指代假设和未解决状态来分配额外的推理过程。一个轻量级的证据评分器（EvidenceScorer）负责评估现有候选对象和未解决状态的置信度。

A training-free Referential Disagreement LOOK policy balances belief-weighted candidate competition and unresolved-region coverage to select a crop for local re-inference. The same benchmark-adapted SAM3-I backbone processes the selected region, and the resulting evidence updates existing hypotheses and introduces new candidates.

一种无需训练的指代差异 LOOK 策略通过平衡置信度加权的候选竞争与未解决区域的覆盖范围，来选择用于局部重推理的裁剪区域。相同的基准适配 SAM3-I 主干网络处理所选区域，产生的证据会更新现有假设并引入新的候选对象。

Among the compared methods, DRA-SAM achieves state-of-the-art performance on six of seven test metrics on both RIS-LAD and UAVid-RIS, including mIoU and P@0.5-P@0.9, and on all seven RIS-LAD validation metrics. On RIS-LAD, DRA-SAM improves the matched global SAM3-I baseline from 49.55 to 52.34 mIoU and from 46.80 to 58.05 P@0.5.

在对比方法中，DRA-SAM 在 RIS-LAD 和 UAVid-RIS 的七项测试指标中的六项上均达到了最先进的性能（包括 mIoU 和 P@0.5-P@0.9），并在所有七项 RIS-LAD 验证指标上表现优异。在 RIS-LAD 上，DRA-SAM 将匹配的全局 SAM3-I 基线从 49.55 mIoU 提升至 52.34，并将 P@0.5 从 46.80 提升至 58.05。

On UAVid-RIS, the independently adapted DRA-SAM achieves 73.26 mIoU and 86.86 P@0.5. Controlled crop-policy comparisons demonstrate the effectiveness of disagreement-guided crop selection, while target-size analysis reveals particularly pronounced gains for aerial targets occupying less than 1% of an image.

在 UAVid-RIS 上，独立适配的 DRA-SAM 达到了 73.26 mIoU 和 86.86 P@0.5。受控的裁剪策略对比证明了差异引导裁剪选择的有效性，而目标尺寸分析显示，对于占据图像不到 1% 的空中目标，其增益尤为显著。

These results establish referential disagreement as an effective criterion for allocating local inference. Code is available at this https URL.

这些结果确立了指代差异作为分配局部推理的有效准则。代码可在该 HTTPS 链接获取。