Abstract: In Synthetic Aperture Radar (SAR) image ship detection, the loss of detailed information and edge blurring caused by coherent speckle noise and low signal-to-clutter ratio often degrade detection accuracy. To address this problem, this paper proposes a ship detection algorithm for SAR images based on the Wavelet Shallow Residual Network (WSSRNet). First, a Multi-scale Wavelet Transform Residual Module (MWTRM) is constructed in the backbone feature extraction network, which introduces wavelet transform convolution to capture diverse receptive fields and fuse multi-scale features, thereby enhancing the attention to contextual and spatial feature information while reducing background interference and computational overhead. Second, a Shallow Skip Residual Module (SSRM) is proposed, which enhances the utilization of positional information and shallow fine-grained features by designing a shallow feature pyramid structure and cross-stage feature fusion, improving the boundary localization precision and recall of small targets. Finally, a novel loss function, Inner-ShapeIoU, is introduced, which adaptively adjusts the regression strategy based on the scale differences of targets, improving the network's generalization performance and boundary localization precision. Experimental results on the High-Resolution SAR Images for Ship Detection (HRSID) dataset show that the proposed method achieves an accuracy of 92.3%, a recall of 87.1%, and a mean average precision (mAP) of 94.7% at an IoU threshold of 0.5, and 70.2% at IoU thresholds of 0.5-0.95. The proposed method demonstrates superior performance in detection accuracy, parameter efficiency, and small target feature extraction capability, with strong multi-scene detection performance and robustness, significantly alleviating the problems of false detection and missed detection for small targets in complex scenarios.