Abstract: To enhance the reliability of γ-ray weld image defect detection in radiographic films, this paper proposes an improved YOLOv5s-based defect detection model, achieving efficient and accurate identification of weld defects in complex environments. In response to excessive channels and redundant information in the convolutional network of the original YOLOv5s model. Firstly, we integrate the SCCONV module into the C3 block of the backbone network, reducing redundancy and boosting detection performance. Secondly, considering the characteristics of weld defects, such as morphological diversity, scale variation, and low contrast, convolutional attention modules (CBAM and SE) are introduced to enhance the attention of the model to the region of interest. Finally, the EIoU loss function replaces the traditional CIoU loss in bounding box regression detection, significantly improving detection accuracy and robustness. Experimental results demonstrate that the enhanced model outperforms the baseline YOLOv5s algorithm across key metrics. Specifically, the accuracy is increased by 4.2%, the recall rate is increased by 3.2%, and the mAP value is in-creased by 3.4%, which verifies the effectiveness of the method in the detection of γ-ray weld defects.