 Our proposed rotation dense feature pyramid networks, RDFPN, is a novel ship detection framework that addresses the limitations of traditional methods. It uses dense connections to enhance feature propagation and reduce redundancies, while also employing a rotation anchor strategy to minimize the number of redundant regions. This results in improved recall and completeness of the semantic and spatial information. Experimental results demonstrate that RDFPN outperforms other state-of-the-art methods in terms of accuracy and efficiency. This article was authored by Shwayang, Haosun, Kunfu, and others.