🤖 AI Summary
This work addresses the challenges of ship detection in synthetic aperture radar (SAR) imagery, which is hindered by speckle noise, complex coastal clutter, and the prevalence of small targets—factors that compromise the robustness and fine-grained feature preservation of conventional optical detectors. To overcome these limitations, the authors propose a domain-aware detection model built upon the DETR framework. The approach introduces two key innovations: a SARESMoE module that employs a sparse gating mechanism to dynamically route features between frequency-domain and wavelet-domain experts, and a Space-to-Depth Enhanced Pyramid (SDEP) that effectively integrates high-resolution shallow spatial features to improve small-target localization. Evaluated on benchmarks including HRSID, the model substantially outperforms YOLO variants and existing SAR-specific detectors, achieving an mAP50:95 of 76.4% and an mAP50 of 93.8%.
📝 Abstract
Ship detection in Synthetic Aperture Radar (SAR) imagery is fundamentally challenged by inherent coherent speckle noise, complex coastal clutter, and the prevalence of small-scale targets. Conventional detectors, primarily designed for optical imagery, often exhibit limited robustness against SAR-specific degradation and suffer from the loss of fine-grained ship signatures during spatial downsampling. To address these limitations, we propose SARES-DEIM, a domain-aware detection framework grounded in the DEtection TRansformer (DETR) paradigm. Central to our approach is SARESMoE (SAR-aware Expert Selection Mixture-of-Experts), a module leveraging a sparse gating mechanism to selectively route features toward specialized frequency and wavelet experts. This sparsely-activated architecture effectively filters speckle noise and semantic clutter while maintaining high computational efficiency. Furthermore, we introduce the Space-to-Depth Enhancement Pyramid (SDEP) neck to preserve high-resolution spatial cues from shallow stages, significantly improving the localization of small targets. Extensive experiments on two benchmark datasets demonstrate the superiority of SARES-DEIM. Notably, on the challenging HRSID dataset, our model achieves a mAP50:95 of 76.4% and a mAP50 of 93.8%, outperforming state-of-the-art YOLO-series and specialized SAR detectors.