Lightweight model for small target detection of SAR images of ships based on NWD loss

Synthetic Aperture Radar (SAR) has the advantages of all-weather and high resolution, and is an effective tool for ship monitoring. SAR image ship detection suffers from high difficulty in small target detection and the existing detection models are complex and computationally intensive. To address these issues, this paper proposes a lightweight model based on YOLOv5, lightweight modules EAM and F-C3 were designed to reduce the computational effort and complexity of the model, The NCBS module is designed and the loss calculation of the model is improved based on NWD to improve the detection accuracy of small targets. Through ablation experiments and model testing, compared with the YOLOv5s model, the model volume is 14% of the original model, the number of parameters (Params) is 11% of the original model, and the FLOPs are 10% of the original model. As shown by the test results, the model detects small targets better than YOLOv5s.