SAR船舶分类的跨数据集迁移与不平衡未知类检测
原标题：Cross-Dataset Transfer and Unknown-Class Detection in Imbalanced SAR Ship Classification
AI 摘要
该研究提出 SAR-Shift-Open Benchmark（SSOB）评估框架，在 FUSARShip 和 OpenSARShip 两个 SAR 船舶数据集上，对 DOFA、Prithvi、ScaleMAE、SSL4EO、SARDet100K、SARJEPA 六个预训练地球观测基础模型进行域内分类、跨数据集迁移和未知类检测三类评测。结果显示域内 SARDet100K 平衡准确率最高（FUSARShip 73.4%、OpenSARShip 53.2%），但跨数据集迁移出现严重失败，部分模型准确率尚可而平衡准确率接近随机（如 64.5% 准确率对应 33.3% 平衡准确率）。未知类检测性能依赖留出类别与数据集，MC-dropout 方差常接近随机，任务特定不确定性分数通常比其更具信息量。
正文节选
Cross-Dataset Transfer and Unknown-Class Detection in Imbalanced SAR Ship Classification Abstract Ship classification from Synthetic Aperture Radar (SAR) imagery is a critical computer vision task, yet the robustness of models under deployment shifts remains unclear. While models are often trained on one dataset and deployed on another, we lack a comprehensive understanding of their cross-dataset generalization. To address this, we evaluate six pretrained models on two SAR ship datasets in three