发表机构
University of Bristol（布里斯托大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出DTKDP框架，通过可学习门控剪枝和双教师蒸馏实现轻量级有向SAR船舶检测，大幅压缩模型并保持精度。
AI 中文摘要
两阶段有向检测器在合成孔径雷达（SAR）船舶检测中实现了高定位精度，但其庞大的骨干网络、特征金字塔、候选区域模块以及沉重的感兴趣区域（RoI）头阻碍了部署。现有的轻量级SAR船舶检测器通常采用单阶段框架，缺乏用于精确旋转定位的候选区域级细化。本文提出了一种用于轻量级有向SAR船舶检测的双教师知识蒸馏与剪枝（DTKDP）框架。DTKDP在卷积层、归一化层和线性层中引入可学习门控，以剪枝卷积通道和RoI头神经元。旋转候选区域对齐（RPA）在共享的教师生成旋转候选区域空间中蒸馏教师和学生预测，而双教师方案结合了来自同质主教师的分类和回归指导以及来自异质辅助教师的互补分类线索。在SAR船舶检测数据集（SSDD）和SAR图像旋转船舶检测数据集（RSDD-SAR）上的实验表明，DTKDP将配备ResNet-50骨干的Oriented R-CNN和RoI Transformer的参数减少了87.5%-91.8%，浮点运算量（FLOPs）减少了75.6%-79.9%。在平均精度（AP）和平均精度均值（mAP）方面，生成的Oriented R-CNN-slim和RoI Transformer-slim保持了接近其全尺寸对应物的精度。在AP50、AP75、mAP50:75和mAP50:95上的相对变化范围从下降2.38%到提升0.65%。与RTMDet-tiny相比，它们在两个数据集上的所有四个指标上提高了0.52%-27.55%，并且持续优于代表性蒸馏方法，展示了良好的精度-效率权衡。
英文摘要
Two-stage oriented detectors achieve high localization accuracy in synthetic aperture radar (SAR) ship detection, but their large backbones, feature pyramids, proposal modules, and heavy region of interest (RoI) heads hinder deployment. Existing lightweight SAR ship detectors typically use one-stage frameworks that lack proposal-level refinement for precise rotated localization. This paper presents a dual-teacher knowledge distillation and pruning (DTKDP) framework for lightweight oriented SAR ship detection. DTKDP introduces learnable gates into convolutional, normalization, and linear layers to prune convolutional channels and RoI-head neurons. Rotated proposal alignment (RPA) distills teacher and student predictions in a shared teacher-generated rotated proposal space, while a dual-teacher scheme combines classification and regression guidance from a homogeneous main teacher with complementary classification cues from a heterogeneous auxiliary teacher. Experiments on the SAR Ship Detection Dataset (SSDD) and Rotated Ship Detection Dataset in SAR Images (RSDD-SAR) show that DTKDP reduces the parameters of Oriented Region-based Convolutional Neural Network (Oriented R-CNN) and RoI Transformer equipped with ResNet-50 backbones by 87.5-91.8% and their floating-point operations (FLOPs) by 75.6-79.9%. In terms of average precision (AP) and mean average precision (mAP), the resulting Oriented R-CNN-slim and RoI Transformer-slim retain accuracy close to their full-scale counterparts. Relative changes across $\mathrm{AP}_{50}$, $\mathrm{AP}_{75}$, $\mathrm{mAP}_{50:75}$, and $\mathrm{mAP}_{50:95}$ range from a 2.38% decrease to a 0.65% improvement. Compared with RTMDet-tiny, they improve all four metrics on both datasets by 0.52-27.55% and consistently surpass representative distillation methods, demonstrating a favorable accuracy-efficiency trade-off.
CommentsPublished in Remote Sensing. The version of record is available at https://doi.org/10.3390/rs18183172
Journal refRemote Sensing 18(18), 3172 (2026)
