Infrared ship detection presents significant challenges due to complex background interference and the sparse semantic information at ship edges. In order to tackle these challenges, we put forward AFEI-Net, an infrared ship detection network based on adaptive feature selection and edge information enhancement. First, to mitigate the effects of complex background interference, we design an Adaptive Feature Selection Module (AFSM), which extracts the most relevant features for the targets while effectively suppressing background noise. Second, to overcome the lack of edge semantic information, we devise an Edge Information Enhancement Module (EIEM), which enhances the focus on the edge information of the target’s shape, thereby minimizing the impact of irrelevant features. Lastly, we propose a Lightweight and Efficient Detection Head (LEDH) that significantly reduces computational resource usage while improving detection accuracy. Experimental results demonstrate that our method effectively detects infrared ship images with complex backgrounds and weak edge semantic information on the ISDD and IRSDSS datasets.