To address the issue of poor detection performance for small and medium-sized dense ships in ship target detection, this paper proposes an improved SSD model based on the single-shot multi-box detector (SSD) algorithm for more accurate detection. The algorithm redesigns anchor boxes to adapt to the ship target detection dataset and integrates SE modules into the VGG network to enhance channel features of the input feature map. Additionally, by incorporating the CBAM module, which is responsible for channel and spatial attention mechanisms, the network's ability to perceive and represent important features is improved. Finally, a feature pyramid module is employed to fuse the six layers of features from the original network. These improvements enhance the model's capability to perceive challenging samples and significantly improve the detection performance of small, occluded, and dense ships. A ship target detection dataset containing five types of ships—cargo ships, cruise ships, fishing boats, inflatable boats, and patrol boats—was collected and organized. Ablation experiments were conducted on the collected dataset, and comparisons with classic network models demonstrated the effectiveness of the added modules.