With the modernization of sericulture, the accurate detection of cocoon pests and diseases has become essential to ensuring a healthy and sustainable silk industry. Traditional manual inspection methods are often inefficient and subjective, limiting their applicability in large-scale production environments. This paper proposes an improved object detection model, YOLOv11-iRMB-SWC, based on the YOLOv11 architecture. The model enhances the original C3K2 module in the backbone network by incorporating two novel components: an Inverted Residual Mobile Block (iRMB) and a Shift-Wise Convolution (SWC) module. These enhancements aim to improve the network’s capability for robust feature extraction in complex scenarios. Experimental evaluations on a custom-built cocoon pest and disease dataset show that the proposed model achieves higher performance in terms of precision, recall, and mean average precision at IoU 0.5, while maintaining a lightweight architecture. The results suggest that YOLOv11-iRMB-SWC offers a promising solution for intelligent cocoon disease detection with strong practical application value.