In SAR image detection, small target ships are susceptible to interference from clutter and noise, making accurate classification and detection challenging. Despite significant progress in this field, there has been a lack of methods specifically adapting to the characteristics of small target ships dynamically. This limitation causes the existing dynamic detectors to equally allocate attention to small objects in simple and complex backgrounds, resulting in poor detection of small objects in complex backgrounds. To address this issue, we propose the SAR Dynamic Feature Adaptive Network (Sea-ShipNet). Firstly, we aggregate semantic information at the shallow feature level, significantly enhancing the feature contrast between small targets and the maritime background. Secondly, we propose a dynamic feature adaptive vector to guide image features to the detection head, paying more attention to small targets within complex backgrounds. We conduct comparative experiments with common methods on two SAR ship datasets, further demonstrating the superiority of our approach in detecting small target ships.