RBFA-Net: A Rotated Balanced Feature-Aligned Network for Rotated SAR Ship Detection and Classification

Ship detection with rotated bounding boxes in synthetic aperture radar (SAR) images is now a hot spot. However, there are still some obstacles, such as multi-scale ships, misalignment between anchors and features, the opposite requirements for spatial sensitivity of regression tasks classification tasks. In order to solve these problems, we propose balanced feature-aligned network (RBFA-Net) where three targeted networks designed. They are, respectively, attention feature pyramid (BAFPN), an anchor-guided alignment (AFAN) rotational (RDN). BAFPN improved FPN, module fusing enhancing multi-level by which can decrease negative impact ship differences. AFAN, adopt convolution layer adaptively align features according anchor solving problem. RDN, task decoupling (TDM) adjust maps, conflict task. addition, L1 loss balance loss. Based on SAR rotation dataset, conduct extensive ablation experiments compare our RBFA-Net eight other state-of-the-art networks. The experiment results show that among networks, makes 7.19% improvement mean average precision compared second-best network.