**7. Conclusions**

This paper proposes a lightweight on-board SAR ship detector called Lite-YOLOv5, which (1) reduces the model volume; (2) decreases the floating-point operations (FLOPs), and (3) guarantees the on-board ship detection without sacrificing accuracy. First, two characteristics are used to obtain a lightweight network, i.e., (1) a LCB module is inserted into the backbone network of YOLOv5 and (2) network pruning is applied to obtain a more compact model. Then, four characteristics are used to guarantee the detection accuracy, i.e., (1) an HPCB module to effectively exclude pure background samples and suppress the false alarms; (2) a SDC method to generate superior priori anchor; (3) a CSA model to enhance the SAR ships semantic feature extraction ability; an (4) an H-SPP model to increase the context information of the receptive field. To evaluate the on-board SAR ship detection ability of Lite-YOLOv5, we also transplanted it to the embedded platform NVIDIA Jetson TX2. Experimental results on the Large-Scale SAR Ship Detection Datasetv1.0 (LS-SSDD-v1.0) show that Lite-YOLOv5 can realize a lighter architecture with a 2.38 M model volume (14.18% of the model size of YOLOv5), on-board ship detection with a low computation cost (26.59% of FLOPs of YOLOv5), and superior detection accuracy (1.51% F1 improvement compared with YOLOv5). We also conducted a large quantity of ablation experiments to verify the effectiveness of the proposed modules. Thus, Lite-YOLOv5 can provide high-performance on-board SAR ship detection, which is of grea<sup>t</sup> significance.

In the future, our works will be as follows:


**Author Contributions:** Conceptualization, X.X.; methodology, X.X.; software, X.X.; validation, T.Z.; formal analysis, T.Z.; investigation, X.X.; resources, X.X.; data curation, X.X.; writing—original draft preparation, X.X.; writing—review and editing, T.Z.; visualization, T.Z.; supervision, T.Z.; project administration, T.Z.; funding acquisition, X.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** No new data were created or analyzed in this study. Data sharing is not applicable to this article. The LS-SSDD-v1.0 dataset provided by Tianwen Zhang is available from https://github.com/TianwenZhang0825/LS-SSDD-v1.0-OPEN (accessed on 6 February 2022) to download for scientific research.

**Acknowledgments:** The authors would like to thank the editors and anonymous reviewers for their valuable comments that greatly improved our manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.
